Deli e able 1.1
S a e o he A and
end-use needs e iew
This p ojec has ecei ed unding om
he Eu opean Union’s Ho izon Eu ope
esea ch and inno a ion p og amme
unde g an ag eemen No 101076963
UK pa icipan s in Ho izon Eu ope P ojec
PHOEBE a e suppo ed by UKRI g an
numbe s 10038897 (The In e na ional
Road Assessmen P og amme – iRAP)
and 10056912 (The Floow)
The sole esponsibili y o he con en o his
documen lies wi h he au ho s. I does no
necessa ily e lec he opinion o he Eu opean
Union. Nei he CINEA no he Eu opean
Commission a e esponsible o any use ha may
be made o he in o ma ion con ained he ein.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
1
Documen Con ol Page
Deli e able numbe 1.1
Deli e able i le S a e o he A and end-use needs e iew
Deli e able e sion 1
Wo k Package numbe 1
Wo k Package Ti le PHOEBE F amewo k - Me hodological and echnical app oach
Due da e o deli e y 30/04/2023
Ac ual da e o deli e y 28/04/2023
Dissemina ion le el Public
Type Repo
Edi o (s) Ami Pooyan A gha i, Eleono a Papadimi iou, TUD
Con ibu o (s)
Ami Pooyan A gha i & Eleono a Papadimi iou, TUD
Apos olos Ziakopoulos & Ma ia Oikonomou, NTUA
A una a Pu a unda, Ch is elle Al Haddad, Mohamed Abouelela, &
Cons an inos An oniou, TUM
Shanna Lucchesi, Monica Olyslage s & James B ad o d, iRAP
An onio Pellice & Ma cel Sala, AIM
Sam Chapman, FLOOW
Ped o Homem de Gou eia & And éia Lopes Aze edo, POLIS
Re iewe (s)
Ana Ma ía Pé ez-Zu iaga, UPV
Ma ko Se o ic, EIRA
P ojec name P edic i e App oaches o Sa e U ban En i onmen s
P ojec Ac onym PHOEBE
P ojec s a ing da e 01/11/2022
P ojec du a ion 45 mon hs
Righ s PHOEBE conso ium
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
2
PM e o s pe bene icia y ha con ibu ed o he deli e able
Documen His o y
# Pa ne PM e o in he Deli e able
1 TUD 2.00
2 NTUA 1.00
3 TUM 3.00
4 iRAP 1.00
5 AIM 2.00
6 FLOOW 0.22
7 POLIS 0.65
8 EIRA 0.03
To al PHOEBE Conso ium
9.9
Ve sion Da e Bene icia y Desc ip ion
0.1 30/01/2023 Eleono a Papadimi iou &
Ami Pooyan A gha i (TUD)
Ou line o he Deli e able
s uc u e and d a
objec i es & me hods
0.2 28/03/2023
Ami Pooyan A gha i &
Eleono a Papadimi iou
(TUD)
Apos olos Ziakopoulos &
Ma ia Oikonomou (NTUA)
A una a Pu a unda,
Ch is elle Al Haddad,
Mohamed Abouelela, &
Cons an inos An oniou
(TUM)
Shanna Lucchesi, Monica
Olyslage s & James
B ad o d (iRAP)
An onio Pellice & Ma cel
Sala (AIM)
Sam Chapman (FLOOW)
Ped o Homem de Gou eia
& And éia Lopes Aze edo
(POLIS)
Inpu o he di e en sub-
chap e s, namely:
Road sa e y assessmen
(iRAP, NTUA)
T a ic mic osimula ion
(AIM)
Human beha iou models
(TUD)
Eme ging da a collec ion
me hods (FLOOW)
Mode choice and modal
shi (TUM)
Socioeconomic impac s
(TUM)
S akeholde s su ey
me hodology & esul s
(NTUA)
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
3
Focus g oups
me hodology and esul s
(POLIS)
0.3 03/04/2023 Ami Pooyan A gha i (TUD)
Fi s comple e d a
including all pa ne s’
con ibu ion, and
commen s / addi ional
inpu eques s om
pa ne s
0.4 06/04/2023
Ami Pooyan A gha i &
Eleono a Papadimi iou
(TUD)
Apos olos Ziakopoulos &
Ma ia Oikonomou (NTUA)
A una a Pu a unda,
Ch is elle Al Haddad,
Mohamed Abouelela, &
Cons an inous An oniou
(TUM)
Shanna Lucchesi, James
B ad o d & Monica
Olyslage s (iRAP)
An onio Pellice & Ma cel
Sala (AIM)
Sam Chapman (FLOOW)
Ped o Homem de Gou eia
& And éia Lopes Aze edo
(POLIS)
Addi ional inpu p o ided
and commen s add essed
by all pa ne s in he
di e en sub-chap e s
0.5 10/04/2023 Eleono a Papadimi iou
(TUD)
Execu i e summa y and
conclusions o he
Deli e able, p oo eading,
consolida ion o inal d a
o in e nal e iew (by
EIRA, UPV)
0.6 26/4/2023
Ami Pooyan A gha i &
Eleono a Papadimi iou
(TUD)
Shanna Lucchesi & Monica
Olyslage s (iRAP)
A una a Pu a unda,
Ch is elle Al Haddad,
Mohamed Abouelela, &
Cons an inos An oniou
(TUM)
Inco po a ion o
commen s om in e nal
e iew and pa ne s
(TUD, iRAP)
Inpu o induced demand
modelling (TUM)
0.7 27/4/2023
Shanna Lucchesi & Monica
Olyslage s (iRAP)
Eleono a Papadimi iou
(TUD)
Final syn hesis wi h
es uc u ing and
igh ening, consolida ion
o key messages and
conclusions
1.0 28/4/2023
Shanna Lucchesi & Monica
Olyslage s (iRAP)
Eleono a Papadimi iou
(TUD)
Final documen
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
4
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
5
P ojec summa y
The EU- unded ‘P edic i e App oaches o Sa e U ban En i onmen ’ (PHOEBE) p ojec aims o de elop
an in eg a ed, dynamic human-cen ed p edic i e sa e y assessmen amewo k in u ban a eas. This will
be achie ed by b inging oge he he in e disciplina y powe o a ic simula ion, oad sa e y assessmen ,
human beha iou , mode shi and induced demand modelling and new and eme ging mobili y da a.
Focused on ulne able oad use s' sa e y, he 3.5-yea -long PHOEBE p ojec will d aw inspi a ion om
eal-wo ld scena ios in he h ee pilo ci ies o A hens (GR), Valencia (ES) and Wes Midlands (UK).
Tes ing ac i i ies will be pe o med ac oss he use cases o simula e and o ecas he impac o changes
on sa e y in di e en scena ios o dis up ions o ansi ions ac oss u ban anspo ne wo ks.
P edic ing and isualising he sa e y and socioeconomic ou comes o new o ms o anspo , new
echnologies, o egula o y and beha iou al changes om he indi idual (mic o) le el up o he ne wo k-
wide (mac o) le el will also be a signi ican game-change o u ban s akeholde s. The esul s o PHOEBE
can be used as a bluep in by o he Eu opean ci ies o de elop hei knowledge p oduc s, such as
socioeconomic analysis model, u ban oad sa e y assessmen , human beha iou and choice modelling.
PHOEBE pilo ci ies
Lis o pa icipa ing ci ies:
• A hens (G eece)
• Valencia (Spain)
• Wes Midlands (Uni ed Kingdom)
Social Links:
h ps:// wi e .com/P ojec _PHOEBE
h ps://www.linkedin.com/company/phoebe-p ojec /
h ps://www.you ube.com/@phoebep ojec
Fo u he in o ma ion please isi WWW.PHOEBE-PROJECT.EU
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
6
P ojec pa ne s
O ganisa ion Coun y Abb e ia ion
EVROPSKI INSTITUT ZA OCENJEVANJE CEST - EURORAP SI EIRA
ETHNICON METSOVION POLYTECHNION EL NTUA
TECHNISCHE UNIVERSITEIT DELFT NL TUD
TECHNISCHE UNIVERSITAET MUENCHEN DE TUM
AIMSUN SLU ES AIM
POLIS AISBL BE POLIS
FACTUAL CONSULTING SL ES FC
UNIVERSITAT POLITECNICA DE VALENCIA ES UPV
OSEVEN SINGLE MEMBER PRIVATE COMPANY EL O7
THE FLOOW LIMITED UK FLOOW
INTERNATIONAL ROAD ASSESSMENT PROGRAMME UK iRAP
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
7
Lis o abb e ia ions and ac onyms
Ac onym
Meaning
AADT
A e age Annual Daily T a ic
AASHTO
Ame ican Associa ion o S a e Highway O icial
ABM
Agen -based Modelling
ADAS
Ad anced D i e Assis ance Sys em
AI
A i icial In elligence
BCA
Bene i Cos Analysis
BCR
Bene i Cos Ra io
CAV
Connec ed and au oma ed Vehicle
CF
Ca Following
CMF
C ash Modi ica ion Fac o
CPI
C ash Po en ial Index
CPM
C ash P edic ion Model
DR
Decele a ion Ra e
EC
Eu opean Commission
EVT
Expec ed Value Theo y
FSI
Fa al and Se e e Inju ies
FHWA
Fede al Highway Adminis a ion
ITF
In e na ional T anspo Fo um
MC
Mode Choice
MCP
Mul i C i e ia P ocess
ML
Machine Lea ning
MNL
Mul inomial Logi
MS
Modal Shi
OD
O igin Des ina ion
PET
Pos Enc oachmen Time
RUM
Random U ili y
SPF
Sa e y Pe o mance unc ion
SRIP
Sa e Roads In es men Plan
SRS
S a Ra ing Sco e
SSAM
Su oga e Sa e y Assessmen Model
TC
Task Capabili y
TTC
Time To Collision
VRU
Vulne able Road Use s
WP
Wo k Package
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
8
Table o con en
1 Pu pose o he deli e able ................................................................................................................ 18
1.1 A ainmen o he objec i es and explana ion o de ia ions .................................................... 18
1.2 In ended audience ................................................................................................................... 19
1.3 S uc u e o he deli e able and links wi h o he wo k packages/deli e ables ........................ 19
2 Me hodology ...................................................................................................................................... 20
2.1 Li e a u e e iew o he s a e-o - he-a ................................................................................... 20
2.2 S akeholde s su ey ................................................................................................................ 29
2.3 Focus g oups .......................................................................................................................... 30
3 Re iew o he S a e o he A ........................................................................................................... 35
3.1 T a ic simula ion, human beha iou and oad sa e y assessmen ......................................... 35
3.2 Modal shi models .................................................................................................................. 69
3.3 Induced demand modelling ..................................................................................................... 78
3.4 Socio-economic impac e alua ion .......................................................................................... 82
3.5 Da a collec ion ools and me hods .......................................................................................... 91
3.5 Discussion on e iew indings ............................................................................................... 100
4 S akeholde consul a ions ............................................................................................................... 109
4.1 O e iew o su ey esul s .................................................................................................... 109
4.2 Focus g oups ......................................................................................................................... 118
4.3 Use Needs S a emen s o he design o he PHOEBE F amewo k .................................... 125
5 Conclusions and ecommenda ions ................................................................................................ 129
5.1 Implica ions on he heo e ical de elopmen o he PHOEBE F amewo k ............................ 129
5.2 Implica ions on he echnical de elopmen o he PHOEBE F amewo k .............................. 131
5.3 Implica ions on he design o end-use ools and knowledge p oduc s ................................. 135
6 Re e ences ...................................................................................................................................... 136
Annex A: Li e a u e e iew guidelines .................................................................................................... 161
Annex B: Addi ional con en o he S a Ra ings, FSIs and SRIP Models .............................................. 163
Annex C: S akeholde s su ey s uc u e ................................................................................................ 170
Annex D: Addi ional s akeholde s su ey esul s g aphs ....................................................................... 185
Annex E: Focus g oup sc ip ................................................................................................................... 197
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
15
collec ion, oad sa e y assessmen , a ic mic osimula ion) conside ing ac o s such as human beha iou ,
modal shi , and imp o ed da a exploi a ion h ough machine lea ning me hodologies. The o al du a ion
o he ques ionnai e is es ima ed a abou 20-25 minu es. Consequen ly, he ques ionnai e is composed
o he ollowing sec ions:
1. Wa m-up ques ions: p o ide a basic anonymous p o ile o be e unde s and he backg ound o
he pa icipan s.
2. Beha iou al models hema ic ques ions: Ra ing he impo ance o a la ge a ay o human
beha iou s ha may eme ge as a esul o changes in anspo sys ems and in luence isk.
3. Da a collec ion me hods hema ic ques ions: Ra ing he impo ance o new ends in da a
collec ion me hods, such as (indica i ely) c owdsou cing, d ones, social media da a, e c.
4. Modal shi hema ic ques ions: Ra ing he impo ance and impac o modal shi o he anspo
ne wo k.
5. Socioeconomic models hema ic ques ions: Ra ing he impo ance and in luence o
socioeconomic pa ame e s o he oad use s, such as a ailable income, unemploymen , age
p o iles, o he ou pu s and he beha iou o he anspo ne wo k.
6. Road sa e y assessmen hema ic ques ions: Ra ing and assessmen o cu en ly a ailable oad
sa e y assessmen me hodologies o ocus on in as uc u e isk and en ich he o he a ailable ools
wi h hem in u n.
7. T a ic mic osimula ion hema ic ques ions: Ra ing he impo ance o enhancing he cu en ly
a ailable a ic mic osimula ion ools.
Closing ques ions: Assessing he ele ance and impo ance o u u e ools de eloped wi hin
PHOEBE o he daily wo k and p o essional p ac ice o pa icipan s.
Addi ionally, he ocus g oups sough o collec inpu , ia a dedica ed sc ip and p ocess, on:
• How local au ho i ies make decisions in he oad sa e y domain, pa icula ly decisions
ega ding managemen and changes o he oad ne wo k, and ulne able oad use s;
• I and how local au ho i ies assess oad sa e y isks in he in as uc u e o Vulne able Road
Use s;
• I and how local au ho i ies use a ic simula ion and/o induced demand modelling,
pa icula ly in he oad sa e y domain;
• I and how local au ho i ies access and use eme ging mobili y da a in mobili y planning and
managemen , and pa icula ly in Road Sa e y;
• Wha kind o capaci y gaps and needs may eme ge wi h he adop ion o a amewo k wi h he
o e all cha ac e is ics o he PHOEBE F amewo k.
The s akeholde s su ey was aken by 50 pa icipan s (41 om Eu ope) om 36 ci ies.
Key messages:
The online s akeholde su ey and ocus g oup consul a ions cap u ed he cu en gaps and needs om
he pe spec i e o anspo manage s and municipali ies ( he in ended end use s o he PHOEBE
F amewo k). These a e syn hesised in o a use needs s a emen o in o m he design o he PHOEBE
F amewo k’s end-use ools and knowledge p oduc s.
The su ey pa icipan s indica ed ha he ollowing scena ios need o be p io i ised: i) implemen a ion o
egula o y measu es o limi speeds; ii) in oducing ex ensi e ne wo k o bicycle lanes; iii) p omo ion o
public anspo modes; i ) in oduc ion o new anspo modes; ) implemen ing hie a chical schemes;
i) encou aging modal shi ii) speed calming measu es and iii) expansion o cycling and walking
in as uc u e.
Rega ding he o e all p o essional needs ele an o PHOEBE p ojec , a eas ha we e iden i ied as high
p io i y o esea ch we e:
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
16
1) T a ic simula ion en iched wi h in as uc u e sa e y in o ma ion, mode shi and induced demand,
and human beha iou al ac o s
2) Imp o ed accu acy o a ic simula ion, and
3) Road assessmen en iched wi h a ic mic osimula ion in o ma ion and AI/ML models.
When asked abou he use ulness o a po en ial ool o e e yday asks, he su ey pa icipan s placed a
high p io i y on he majo aspec s planned unde he PHOEBE p ojec , namely a ic simula ion en iched
wi h in as uc u e sa e y in o ma ion, modal shi in o ma ion, oad assessmen enhanced wi h AI/ML
models. I is in e es ing o no e ha he esponden s om he PHOEBE use case loca ions in A hens,
Valencia and Wes Midlands e ealed some a iabili y among he needs and p io i ies, suppo ing he
no ion ha he PHOEBE F amewo k needs o able o lexible enough o add ess di e en needs.
The ocus g oups ga he ed 19 pa icipan s om 15 di e en ci ies.
Key messages:
• Road Sa e y Unde s anding, which in luences many cu en p ac ices and challenges aced
by p ac i ione s. P o essionals alked abou he di e en ways and pe spec i es o oad sa e y,
wi h “issues on how we app oach oad sa e y”.
o Comp ehensi e cos -bene i analysis should b ing isibili y o he social cos s and
he economic and heal h bene i s
o Shi in socie y as ega ds oad sa e y pe cep ions; c ashes should no be seen
as a “no mal” pa o pa icipa ing in a ic.
• O ganisa ional s uc u es (knowledge and p ac ices) di e among ci ies, and he e is need
o emo e he ‘silos’ be ween di e en ypes o anspo p o essionals, especially wi h espec
o sa e y.
o Main de iciencies in he cu en ools a e he limi ed inco po a ion o modal shi
op ions (especially wi h new eme ging mobili y solu ions) and he
o VRUs and new mobili ies need o be in eg a ed no only in ools bu also wi hin he
o ganiza ional knowledge and p ac ice.
• O ganisa ional esou ces and capabili ies, including he da a aspec , will in luence he
possible implemen a ion o new amewo ks wi hin hese au ho i ies and o ganisa ions.
o The new eme ging da a sou ces a e cos ly o collec and p ocess, and he lack o
human esou ces wi h ele an aining and knowledge may be a ba ie in he
exploi a ion o hese new da a.
• Concep s eali y: a clea e unde s anding is needed o he link be ween design guidelines
and hei ac ual sa e y impac s, and he alidi y o he espec i e modelling ou pu s.
The esul s o he s akeholde s su ey and he ocus g oups a e summa ized in he o m o a Needs
S a emen on h ee elemen s: i) s a egic goals; ii) decision suppo and daily p ac ice; and iii)
me hodological needs.
Conclusions
On he basis o s akeholde s’ needs, as well as he in-dep h e iew o he scien i ic and g ey li e a u e
o e he PHOEBE F amewo k componen s, a concep ual design o he PHOEBE me hodological
amewo k will be d a ed on he basis o he ollowing design p inciples:
i) Fusion o oad assessmen and a ic simula ion: he s ong in e - ela ionship be ween
a ic simula ion and oad sa e y assessmen is a he co e o he PHOEBE me hodology.
ii) Inco po a ion o human beha iou models o enhance he c edibili y o a ic simula ion and
en ich he p edic ions o oad sa e y ou comes
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
17
iii) Mode choice and modal shi models a e he i s s ep o be aken in o de o simula e
a ic and sa e y impac s.
i ) Socio-economic impac s a e he inal ou comes o be es ima ed om he a ic and sa e y
ou pu s o he amewo k.
) The new eme ging da a sou ces a e a e ical dimension ha will enhance he p edic i e
and explana o y powe o all componen s o he amewo k.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
18
1 Pu pose o he deli e able
1.1 A ainmen o he objec i es and explana ion o de ia ions
This deli e able, which is he di ec ou pu o Task 1.1 o he p ojec , aims o e iew he s a e-o - he-a in
each componen u ilised in he PHOEBE F amewo k and gain a deepe unde s anding o he needs o
ci ies and anspo planne s.
Fi s , a li e a u e e iew highligh s he gaps (in e ms o bo h esea ch and p ac ice) in he exis ing
me hodologies and echnologies ha a e ele an o PHOEBE and sheds mo e ligh on he inno a ion and
signi icance o PHOEBE in how i goes beyond he s a e-o - he-a . The li e a u e e iews a e documen ed
by he espec i e echnical pa ne s (NTUA and iRAP o oad sa e y assessmen me hodologies and
ools, AIMSUN o a ic simula ion so wa e and models, TUD o human beha iou al models, TUM o
mode shi and induced demand and socio-economic models, and FLOOW o da a collec ion and
analy ics).
Second, a needs s a emen is de eloped based on s akeholde consul a ions wi h he po en ial end-use s
o he PHOEBE F amewo k. The cu en gaps and needs om he pe spec i e o anspo manage s and
municipali ies ( he in ended end use s o he PHOEBE F amewo k) a e cap u ed in a needs s a emen o
in o m he design o he PHOEBE F amewo k. This was done ia in e iews and su eys wi h s akeholde s
in ol ed in use case demons a ions (led by NTUA, FC, UPV and FLOOW), as well as wi h he POLIS
s akeholde g oup and h ough he Eu oRAP ne wo ks (led by POLIS and EIRA). The implica ions o hese
indings on he de elopmen he me hodological amewo k in PHOEBE a e hen discussed.
The objec i es ela ed o his deli e able ha e been achie ed in ull and as scheduled. Table 1.1 below
shows he dis ibu ion o wo k among pa ne s in his deli e able.
Role Who
D a ing he deli e able, coo dina ing he li e a u e e iews, consolida ing
con ibu ions om all pa ne s, e iewing s a e-o - he-a in beha iou al models,
w i ing he conclusions
TUD
Leading he s akeholde su eys, d a ing he esul s o
s akeholde su eys,
e iewing s a e-o - he-a in oad sa e y assessmen
NTUA
Re iewing s a e-o - he-a in modal shi and socio-economic analysis TUM
Re iewing s a e-o - he-a in oad sa e y assessmen and eme ging da a collec ion
ools and me hods
, con ibu ing o he ocus g oups no e aking, e iewing he
s akeholde s' su ey ques ionnai e
iRAP
Re iewing s a e-o - he-a in a ic mic osimula ion AIM
Re iewing s a e-o - he-a in eme ging da a collec ion ools and me hods FLOOW
Designing and implemen ing ocus g oups, d a ing he esul s o ocus g oups POLIS
Table 1-1 Con ibu ions o PHOEBE pa ne s o he deli e able
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
19
1.2 In ended audience
The in ended audience o his deli e able is he PHOEBE p ojec pa ne s in ol ed in he asks o WP1 –
Me hodology, he anspo manage s and municipali ies in ol ed in he PHOEBE use cases, and b oade
s akeholde g oups in ol ed in u ban anspo and sa e y managemen , namely hose o he POLIS,
Eu oRAP and EIRA ne wo ks.
1.3 S uc u e o he deli e able and links wi h o he wo k
packages/deli e ables
This deli e able will in o m he selec ion o ools and models o be u ilized and de eloped o he PHOEBE
F amewo k by he espec i e echnical pa ne s. Ou pu s will di ec ly in o m he p oposi ion o dynamic
isk assessmen and socio-economic analysis me hodologies (Tasks 1.2 and 1.3) and he speci ica ion o
he echnical aspec s o be de eloped o demons a ion o he PHOEBE F amewo k (Task 1.4).
The deli e able is s uc u ed in o six chap e s as ollows:
• Chap e 1 ou lines he pu pose and aims o Task 1.1
• Chap e 2 desc ibes he me hodology used o he li e a u e e iew and o he iden i ica ion
o end-use needs
• Chap e 3 p o ides he indings o he li e a u e e iew, including exis ing gaps and
implica ions o PHOEBE F amewo k
• Chap e 4 p o ides he indings om s akeholde su eys and ocus g oups, and
• Chap e 5 p o ides he o e all conclusions om he s a e-o - he-a e iew and he end-use
needs, and he implica ions o PHOEBE’s heo e ical and echnical aspec s and end-use
ools and suppo ma e ials.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
20
2 Me hodology
In his chap e , he me hodology implemen ed in PHOEBE o he pu poses o achie ing he objec i es o
his deli e able is desc ibed. The me hodology has h ee dis inc componen s:
1. Re iew o he s a e-o - he-a o documen he body o knowledge a ailable o each
componen o he p edic i e sa e y assessmen amewo k o be de eloped, and d aw
conclusions abou he po en ial implica ions o be aken in o accoun in he design o his
amewo k.
2. Online s akeholde s’ su ey o unde s and he needs and gaps in cu en p ac ice among he
s akeholde s in ol ed in he PHOEBE use cases (A hens, Valencia, Wes Midlands), as well as
b oade g oups o anspo manage s eachable h ough he wide ne wo k o con ac s o he
p ojec pa ne s. Fo ha pu pose, a dedica ed on-line su ey was designed and dispa ched.
3. Focus g oups consul a ion o d ill mo e deeply in o he cu en p ac ices o oad sa e y
assessmen in u ban a eas and he use o mic osimula ion he ein, and he ex en o which hey
in o m decision making in u ban oad planning and managemen .
The syn hesis o he esul s o hese componen s we e used o d a a use needs s a emen and he
clea iden i ica ion o gaps in e ms o bo h scien i ic knowledge and p o essional p ac ice, e en ually
in o ming he de elopmen o he PHOEBE p edic i e assessmen amewo k me hodology. In he
ollowing sec ions, he speci ic me hodological and design app oaches pe componen a e desc ibed in
de ail.
2.1 Li e a u e e iew o he s a e-o - he-a
The s a e-o - he-a in he main componen s o PHOEBE (i.e. oad sa e y assessmen , a ic
mic osimula ion, human beha iou al models, modal shi , induced demand, socioeconomic analysis, and
da a collec ion echnologies and me hodologies) we e e iewed in o de o highligh he gaps ( om
academic and om oad manage s pe spec i es). The main pu pose o he li e a u e e iew was o p o ide
an o e iew o wha exis s in he body o knowledge in o de o selec app op ia e ools and me hods o
de eloping he me hodological amewo k o PHOEBE.
To achie e he abo e objec i es, he s a e-o - he-a was e iewed o each componen o PHOEBE om
academic and g ey li e a u e on he basis o a dedica ed e iew guideline documen (see Annex A). The
ollowing sea ch engines a e used o he e iews:
• Google schola , Scopus o academic li e a u e
• Google o g ey li e a u e (Wo ld Bank, ITF, iRAP, e c.)
• CORDIS websi e o p e ious Eu opean p ojec s
Since each componen in PHOEBE is a b oad opic by and o i sel , he e iews a e na owed down o he
in e sec ion o he opics e.g. oad sa e y assessmen and a ic mic osimula ion. Since he e a e a
numbe o e iews abou he PHOEBE componen s a ailable, p io i y was gi en o ecen e iew a icles
o documen s. The ollowing sea ch s a egy and selec ion c i e ia a e applied o he pu pose o he e iew
in his deli e able:
• Keywo d #1 [example: oad sa e y assessmen ]
• Keywo d #1 AND Keywo d #2 [example: oad sa e y assessmen AND a ic mic osimula ion]
• Keywo d #3 [example: human beha iou ]
• Keywo d #3 AND Keywo d #1 [example: oad sa e y assessmen AND human beha iou ]
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
21
• Keywo d #3 AND Keywo d #1 AND Keywo d #2 [example: oad sa e y assessmen AND
human beha iou AND a ic mic osimula ion]
The s udies we e selec ed based on he ype o s udy, numbe o ‘hi s’, publica ion yea , epu a ion o he
academic jou nal, and popula i y o he p ojec .
Fo quali y assu ance pu poses, he li e a u e a ound di e en opics in PHOEBE was e iewed in pai s
o pa ne s. Fo each subjec , wo pa ne s we e pai ed o sea ch and selec he s udies sepa a ely, which
was ollowed by a ‘consensus’ mee ing o he inal sou ces selec ed o be included in he e iew. The
able below shows he opics o li e a u e e iew and he dis ibu ion o pa ne s ac oss hose opics:
Pai ed pa ne s
Re iew opic
AIM and TUM
Re iew o he exis ing a ic mic o-simula ion me hods and ools
iRAP and NTUA
Re iew o he exis ing oad sa e y assessmen me hodologies
TUD and NTUA
Re iew o he exis ing beha iou al models ( ocus on VRUs)
TUM and TUD
Re iew o he exis ing socio-economic analysis and modal shi modelling
FLOOW and iRAP
Re iew o he exis ing and eme ging da a collec ion ools and me hods
Table 2-1 Dis ibu ion o opics
The ele an s udies we e s o ed in a i ual space ha is sha ed among all pa ne s. These s udies we e
all documen ed in a sha ed able wi h i le, au ho e c. and ags (co esponding wi h he opic o ha s udy
wi hin he li e a u e e iew) o each selec ed s udy.
2.1.1 Li e a u e sea ch s a egies
The li e a u e e iew s a egies employed by each se o pa ne s o iden i y app op ia e li e a u e o e iew
is desc ibed below. In o al, 325 pape s we e iden i ied and e iewed.
2.1.1.1 T a ic mic o-simula ion me hods and ools
The li e a u e sea ch s a egy ela ed o a ic mic o-simula ion me hods and oo s a ed wi h a i s sea ch
in Scopus o academic li e a u e o he ollowing keywo ds: “mic osimula ion”, “ a ic” and “sa e y”. This
p oduced a o al o 410 hi s. The same sea ch was done using Google Schola bu , in his case, mo e
han 19,500 esul s we e ound. A his s age, i was necessa y o na ow down he Google Schola sea ch
and u he concep we e added o he p e ious h ee keywo ds in he e iew. The i s main building
blocks o mic oscopic simula ion models a e ‘ca - ollowing’ models and ‘lane-changing’ models, bu in his
case paying an ex a a en ion in o ‘sa e y’ o mic osimula ion. The e o e, ano he sea ch was done using
hese ex a wo ds, i s ‘lane-changing’ esul ing in 6,200 hi s and second ‘ca - ollowing’ esul ing in 7,810
hi s.
Ano he sea ch was done o e iew he li e a u e ela ed o he impac analysis on ulne able oad use s.
In his case, ‘pedes ian modelling’ was also added as an ex a keywo d o he sea ch, esul ing in a o al
o 307 hi s (conside ing he h ee base keywo ds and his ex a one). The keywo ds ‘cyclis modelling’ and
‘bicycle’ we e also added in ano he sea ch, esul ing in 2,080 hi s.
The gene al sea ch engine Google was also used o sea ch o any ‘g ey’ li e a u e ha we e no included
in Google Schola o Scopus. The e iewe s ha e also included app oaches/ esul s o e en jus
e e ences om a ew Eu opean p ojec s ha ha e al eady inished o a e s ill on-going.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
22
As obse ed, he numbe o pape s ound in he di e en sea ches we e s ill oo high. The e o e, in his
case, a non-sys ema ic e iew was done, by picking some o he mos ele an pape s om he e iewe s’
c i e ion. On op o he p e ious while e iewing pape s, some in e es ing e e ences appea ed. Thus,
some pape s we e no selec ed by a keywo d s a egy bu a he om a ‘snowballing’ om exis ing pape s.
Finally, some known e e ences by he e iewing eam we e added as well. As a esul o his app oach a
o al o 67 pape s ha e been e iewed. A summa y o he li e a u e sea ch s a egy (keywo ds and
Boolean ope a o s) is included in he Table 3.2.
Keywo d (OR)
Boolean
ope a o
Keywo d
Boolean
ope a o
Keywo d
T a ic
Ca - ollowing
Lane-changing
Ca - ollowing AND human ac o s
Lane-changing AND human ac o s
Vulne able Road Use s
Pedes ian modelling
Cyclis modelling
Bicycle
Au oma ed ehicles
AND
Mic osimula ion
OR
Mic oscopic
simula ion
AND
Sa e y
Sa e y
assessmen
Table 2-2 G oup o keywo ds o he li e a u e sea ch
2.1.1.2 Road sa e y assessmen
The li e a u e sea ch ela ed o he exis ing oad sa e y assessmen me hodologies used he sea ch
engines Scopus, Google Schola and he CORDIS websi e o Eu opean p ojec s. The sea ch was
conduc ed in wo pa s. The i s aimed o iden i y exis ing oad sa e y assessmen me hods ocused on
in as uc u e isk. Since oad sa e y assessmen is a b oad bu consolida ed opic, he sea ch ocuses on
e iew pape s ha could speed up he e iew p ocess. The e o e, “Road sa e y assessmen ” o “ oad isk
assessmen ” plus “in as uc u e” plus “ e iew” we e he keywo ds used. We selec ed s udies published
om 2018-2023; he sea ch was pe o med by he end o Janua y 2023. Based on hese c i e ia, he
Scopus sea ch e u ned 54 hi s, while Google Schola had 83 hi s. Ti les and abs ac s o hese hi s we e
e iewed. Fo he CORDIS esea ch, h ee p ojec s we e iden i ied as po en ial sou ces o in o ma ion o
PHOEBE. The pape s and p ojec s selec ed a e discussed in he esul s.
The second sea ch ocused on he in e ac ion among isk assessmen s and simula ion. The sea ches on
Scopus and Google Schola e u ned 126 hi s. O hese, 40 eco ds we e excluded a e e iewing bo h
i les and abs ac s and i was ound ha did no mee he inclusion c i e ia. In addi ion, ou s udies we e
disca ded because he ull ex was no a ailable and h ee addi ional pape s we e no in English.
The e o e, he ull- ex o he emaining 79 eco ds was examined in mo e de ail and 57 s udies we e
ound o mee he c i e ia o be included in he sys ema ic e iew. Two ele an s udies we e also added
by checking he e e ences o he iden i ied pape s and sea ching o s udies ha ha e ci ed hem,
esul ing in 59 pape s in o al.
2.1.1.3 Human beha iou models
The p e ious s udies a ound beha iou al models in anspo science a e ex ensi e. A lo o hese s udies
a e ela ed o a ic psychology and s em om psychological and beha iou al science. Howe e , and o
he pu pose o PHOEBE, he li e a u e e iew a ound beha iou al models in PHOEBE is limi ed o hose
s udies ha ocused on beha iou al models o he pu pose o oad sa e y in a ic mic osimula ion. In
addi ion, oad use (human) beha iou and human ac o s ha e been used in he anspo science
li e a u e in e changeably. The e o e, he li e a u e sea ch included he la e e m as well. Mo e
speci ically, he sea ch s a egy (keywo ds and Boolean ope a o s) o beha iou al models is as in he
Table 3.7.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
23
Keywo d
Boolean
ope a o
Keywo d
Boolean
ope a o
Keywo d
Beha iou al (beha iou al)
model
Human beha iou
(beha iou )
Human ac o s
D i e beha iou (beha iou )
Pedes ian beha iou
(beha iou )
Cyclis (cycling) (bicycle)
beha iou (beha iou )
AND
T a ic
Mic osimula ion
AND
Sa e y
Su oga e sa e y
Sa e y
assessmen
Table 2-3 Li e a u e e iew sea ch s a egy o beha iou al models in PHOEBE
The abo e keywo ds and Boolean ope a o s we e used in Google Schola and Scopus o ind ele an
academic s udies. In doing so, e iew a icles we e p io i ized o ex ac ing key s udies in his ield and
c oss e e encing. In addi ion, he same sea ch s a egy was used o ind ele an epo s and Eu opean
p ojec s in CORDIS. Finally, he same sea ch s a egy was used in Google o ind he ele an “g ey”
li e a u e om uni e sal o ganiza ions such as he Wo ld Bank, ITF, and iRAP.
The abo e sea ch esul ed in a o al o abou 200 academic a icles among which a ound 50 a icles we e
selec ed based on hei i le, abs ac , numbe o ci a ions and jou nal epu a ion. A e ca e ully sc eening
he main ex o hese 50 a icles, only 26 a icles we e ound o be ele an o PHOEBE and hus hey
we e e iewed in de ail. In addi ion, he sea ch esul ed in a o al o 25 Eu opean p ojec s and epo s,
among which 4 we e selec ed as ele an o PHOEBE. No g ey li e a u e was ound a ound beha iou al
models in a ic mic osimula ion.
2.1.1.4 Modal shi
Modelling he Mode Choice (MC) phenomenon and o ecas ing he Modal Sha e and Modal Shi (MS) is
in eg al o modelling a el demand. I ac s as a basis o mul iple decisions made by policymake s.
Al hough modelling MC is a ela i ely new ield, se e al scien i ic s udies ha e been conduc ed, esul ing
in his ield's exponen ial de elopmen . A basis o conduc ing a s uc u ed li e a u e sea ch in his domain
is based on a e iew ques ion men ioned below (A kinson & Cip iani, 2018):
Wha a e he commonly used modal shi o ecas ing me hods?
The e iew ques ion was a basis o gene a ing he ele an keywo ds o a s uc u ed li e a u e sea ch.
The chosen keywo ds a e:
Mode choice, Model, Random U ili y, Machine Lea ning, A i icial In elligence, Modal Shi , Passenge
T anspo
Boolean ope a o s we e used o combine hese keywo ds o gene a e consis en exp essions con o ming
o he e iew ques ions. These exp essions we e used o sea ch o academic s udies and g ey li e a u e
in Scopus and Google schola . The sea ch exp essions a e lis ed below:
Google Schola : mode choice model * " andom u ili y" "machine lea ning" "a i icial in elligence" - eigh
Scopus: TITLE-ABS-KEY ( ( mod* ) AND ( choice OR shi ) AND ( model* OR o ecas * ) AND (
me hod* ) AND ( passenge AND anspo * ) AND ( ( machine AND lea n* ) OR ( andom AND
u ili y ) OR ( a i icial AND in elligence ) ) )
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
24
The sea ch was ca ied ou on 6 h Ma ch 2023. The Scopus da abase ende ed 51 hi s, whe eas he
Google Schola yielded 420 hi s. The sea ch was ollowed by de eloping selec ion c i e ia men ioned
below o il e ou he ele an s udies. The il e ing c i e ia a e:
• S udies in pee - e iewed jou nals OR con e ence p oceedings OR scien i ic p ojec epo s.
• S udies conduc ed in English OR Ge man.
• S udies in es iga ing he MC modelling me hodologies OR MS o ecas ing me hodologies.
Finally, 50 ele an s udies we e chosen o conduc he li e a u e e iew.
2.1.1.5 Induce demand
The s a e-o - he-a e iew o induced demand modelling s a s by b ie ly in oducing he opic. Acco ding
o Wikipedia, induced demand, o la en o gene a ed demand, is he phenomenon ha occu s o appea s
wi h he inc ease in supply. I esul s in a decline in p ice and an inc ease in consump ion. In a ic and
anspo a ion, induced a el demand is in e p e ed as an inc ease in he a el demand i.e., he numbe
o ips due o imp o emen s in he anspo supply sys ems, e.g., oad in as uc u e imp o emen and
capaci y inc ease (Li man, 2017). These de ini ions a e u ilised o de elop a e iew ques ion as a
ounda ion o he li e a u e sea ch ela ed o he induced demand modelling. The e iew ques ion is:
• Wha is induced a el demand, and how is i modelled?
Based on he e iew ques ion, ele an keywo ds a e mapped, which a e used o conduc he li e a u e
sea ch. The keywo ds a e:
Induced a el demand, La en a el demand, conges ion, in as uc u e, ebound, model,
modelling.
These keywo ds a e combined wi h boolean ope a o s o gene a e consis en exp essions ma ching he
e iew ques ion. These exp essions a e used o sea ch o he academic and g ey li e a u e in he Google
Schola da abase. The sea ch exp essions a e as ollows:
Google Schola : model * conges ion in as uc u e ebound "induced a el demand" OR "la en a el
demand"
The sea ch was ca ied ou on 6 h Ma ch 2023. The Google Schola da abase gene a ed 155 hi s. The
sea ch was ollowed by de eloping selec ion c i e ia men ioned below o il e ou he ele an s udies.
The il e ing c i e ia a e:
• S udies in pee - e iewed jou nals OR con e ence p oceedings OR scien i ic p ojec epo s.
• S udies conduc ed in English OR Ge man.
• S udies which in es iga e and app aise he socio-economic impac s o oad sa e y measu es.
Finally, 23 ele an s udies a e chosen o conduc he li e a u e e iew. The s udies common o he a el
demand modelling a e no chosen explici ly as hey a e al eady conside ed in he e iew o he mode
choice modelling (session 2.1.1.4).
2.1.1.6 Socio-economic analysis
The s a e-o - he-a e iew s a s wi h he de elopmen o a li e a u e sea ch s a egy. The e ms
"Socioeconomics" and "Socio-economic assessmen o analysis" a e o mally de ined as a basis o he
s a egy de elopmen .
Acco ding o Hellmich (2017), socioeconomics is an in e disciplina y ield o s udy ha simul aneously
ocuses on he in e dependencies be ween socie y's economic and social ac o s. I also seeks o p o ide
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
31
2.3.1 Focus g oup ec ui men , acili a ion and epo ing
A a ge g oup was iden i ied o app op ia ely answe he abo e-men ioned ques ions and ul il he
objec i es se o he ocus g oups. This a ge g oup ep esen ed he cu en p o essionals in he anspo
planning, a ic managemen and oad sa e y sec o s, including hose wo king wi hin and o public
au ho i ies. The ollowing c i e ia we e used:
a) P o essionals wo king o local au ho i ies in he anspo sec o ;
b) A leas wo yea s o expe ience in anspo planning OR a ic managemen OR oad sa e y;
c) Some p esen OR pas expe ience in he use o a ic simula ion (no necessa ily as a modelle ).
POLIS used i s ne wo k, and especially he ne wo k om he Road Sa e y Wo king G oup, o each ou o
p o essionals and ec ui pa icipan s. POLIS Road Sa e y Wo king G oup con ac s include POLIS
membe s and non-POLIS membe s ha wo k in Eu ope and beyond, co e ing a conside able a ie y o
p o essionals wo king in di e en aspec s o oad sa e y and wi h good geog aphical co e age.
P ospec i e pa icipan s we e o e ed a ound six di e en imeslo s o choose om. The ocus da es and
imes a e shown in Table 2.2. The ocus g oup sessions we e conduc ed online wi h a du a ion o 1h30m
and English being he p ima y language and no mo e han six pa icipan s pe session. Each session had
one acili a o and wo no e- ake s. Deb ie ing sessions ook place igh a e all pa icipan s logged o he
online mee ing. The deb ie ing sessions included (1) o e all imp essions o he session, (2) highligh s o
he discussion wi h pa icula ele ance o he esea ch goals, (3) p io i isa ion o u he explo a ion o
speci ic opics o angles in subsequen sessions, (4) assessmen o he acili a ion and (5) e en ual
adjus men s o he sc ip , namely in e ms o g oup dynamics and eamwo k. Facili a ion was done by he
POLIS Road Sa e y Wo king G oup coo dina o , and POLIS and iRAP did no e aking.
In o al, POLIS conduc ed i e ocus g oups, wi h 19 pa icipan s om 11 di e en coun ies and 15
di e en ci ies.
Focus g oup pa icipan s we e in i ed o pa icipa e unde he ollowing p i acy condi ions:
a) The epo and any o he ma e ial o be dissemina ed beyond he s ic ci cle o PHOEBE pa ne s
in ol ed and will no men ion hei names and ins i u ions;
b) Thei inpu s may be quo ed ipsis e bis, bu nei he he con en no he a ibu ion will be made in
a way ha can enable iden i ica ion o he au ho o ins i u ion.
Following his p i acy commi men , he no es and he sound o ideo eco dings aken o he ocus g oup
sessions a e being s o ed in passwo d-p o ec ed s o age (as de ailed in p ojec PHOEBE’s da a
managemen plan D7.2 and da a egis e s D1.2), and accessed and used exclusi ely o his p ojec . The
sound and ideo eco dings will be elimina ed a e he ull conclusion o WP1, including app o al o
espec i e deli e ables by he Eu opean Commission’s p ojec o ice .
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
32
Pa icipan s Coun y 10.Feb -
11h00 o
12h30
15 Feb -
11h00 o
12h30
15 Feb-
14h00 o
15h30
17.Feb -
11h00 o
12h30
20.Feb -
15h30 o
17h00
P01
USA
-
-
-
-
1
P02
Po ugal
1
-
-
-
-
P03
Japan
1
-
-
-
-
P04
Mexico
-
-
1
-
-
P05
Spain
1
-
-
-
-
P06
No way
-
1
-
-
-
P07
Po ugal
-
-
-
1
-
P08
La ia
-
1
-
-
-
P09
Po ugal
-
-
1
-
-
P10
Po ugal
-
-
-
-
1
P11
Po ugal
-
1
-
-
-
P12
Ne he lands
-
-
1
-
-
P13
Po ugal
1
-
-
-
-
P14
Ge many
1
-
-
-
-
P15
Poland
-
1
-
-
-
P16
Po ugal
-
-
-
1
-
P17
Po ugal
-
-
-
1
-
P18
Is ael
-
-
1
-
-
P19
Ge many
-
-
-
1
-
Table 2-6. Focus g oup cha ac e is ics (coun y o o igin, da e and ime)
2.3.2 Sc ip
The ocus g oups ollowed a sc ip ha was de eloped o add ess he objec i es desc ibed abo e and in
cohe ence wi h he a ge g oup.
Pa ne s om WP1 p o ided inpu and eedback o he sc ip be o ehand. None heless, a e each ocus
g oup and deb ie ing session, necessa y adjus men s o made o imp o e he acili a ion o he ocus
g oups and he discussion. The inal de ailed sc ip is in Annex E: Focus g oup sc ip
Below a e he main ques ions om he ocus g oups:
Q1 - Wha ’s he impo ance o Road Sa e y o he wo k you’ e doing?
• How is i impo an ?
• I i should be impo an bu isn’ – why is ha ?
• Wha kind o decisions a e made
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
33
• Balancing o sa e y wi h o he conce ns (e.g., conges ion)
• The ole o poli ical powe , decisions
Q2 - Se e al decisions can ha e an impac on Road Sa e y. A di ec o indi ec impac . A posi i e o
nega i e impac . Special decisions, bu also e e yday decisions. Decisions made by you – and also by
you colleagues and leade s.
You’ e ce ainly seen o he people (colleagues, poli ical leade s, clien s) make hose ypes o decision.
Wha a e he issues hey usually conside as impo an in hose decisions?
• Conce ns o o he s: cos , conges ion, poli ical p essu e
• Types o decisions made
• Use beha iou
• Me hod – desc ip ion, imp o emen , needs o imp o e
• T ansposi ion o he me hod o he scale o he o ganisa ion
Q3 - People walking, cycling, o using e-scoo e s a e usually called “ ulne able oad use s”. Because hey
a e conside ed o be a a highe isk han ca d i e s.
How do you e alua e he isk o hese ulne able oad use s?
• P ac ices o he OTHERS inside o ganisa ion
• Role o in as uc u e on beha iou
• T a ic olumes, Speed limi s, P ac iced speed
• Role o in as uc u e on beha iou
• Me hods o assess isk
• Wha kind o isky beha iou s a e hey in e es ed in?
• Beha iou al ac o s ha a ec oad sa e y
Q4 – Some local au ho i ies use a ic simula ion o suppo anspo planning o managemen . The use
o a ic simula ion so wa e implies building a digi al model o eali y, in oducing a ic da a, and seeing
wha a e he implica ions ha some changes may ha e on a ic beha iou . The simula ion p oduces
esul s – indica ing, o example, i he e will be mo e o less a ic conges ion, how a ic will be
edis ibu ed, e c.
Imagine you could o de a new a ic simula ion so wa e. You could o de wha ypes o esul s his
so wa e would p o ide. Wha ypes o esul s would be use ul o people wo king on Road Sa e y?
• Used, yes o no? Would like o? Wha could help YOU use i ?
• Did you ind i p ac ical? Did you ind i use ul?
• Wha a e he obs acles o using i – o using i mo e?
• P ac ical di icul ies wi h using a ic simula ion
• O ganisa ional di icul ies – s a , s a ime, budge , IT, da a
Q5 - In he pas ew yea s se e al new mobili y se ices a i ed on ou s ee s. Digi al echnology and
geoloca ion play an impo an ole. These se ices gene a e la ge amoun s o digi al da a. These da a can
be use ul o mobili y planning o managemen .
Do you hink hese da a can be use ul o Road Sa e y? How?
• O he ypes and sou ces o da a
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
34
• Di icul ies in acqui ing da a
• In e nal o ganisa ional ba ie s o access o use o he da a
• Lack o echnical skills (knowledge, expe ience)
• Lack o means (so wa e, e c.)
Q6 – We a e wo king on a p ojec ha will b ing oge he new echniques, new da a sou ces, and new
knowledge o ad ance Road Sa e y. Implemen ing a new me hod and ools o his kind in an o ganisa ion
is usually a challenge. We’ e looking o he bes ways o make i easy o local au ho i ies and hei
p o essionals o adop hese solu ions. Imagine we we e o b ing o you o ganisa ion [o a public au ho i y
you ha e wo ked o o wi h] a new me hod and ool o suppo decisions a ec ing Road Sa e y. Imagine
his new me hod and ool b ings oge he new echniques, new da a sou ces, and new knowledge.
Wha kind o di icul ies would you expec in you o ganisa ion?
• Wha could help o e come hose di icul ies?
• Wha ad ice do you ha e o us?
• Reques o de ails abou he amewo k – wha a e hey asking?
• P ac ical ad ice
• P e ious di icul ies – own o o OTHERS
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
35
3 Re iew o he S a e o he A
The main componen s o he PHOEBE F amewo k a e oad sa e y assessmen , a ic mic osimula ion,
human beha iou al models, modal shi , induced demand, socioeconomic analysis, and da a collec ion
echnologies and me hodologies. Du ing Task 1.1 he s a e-o - he-a o each o hese componen s was
e iewed. This chap e p esen s he indings o his e iew o each componen . The chap e concludes
wi h a discussion and key conside a ions a ising om he e iews.
3.1 T a ic simula ion, human beha iou and oad sa e y assessmen
T a ic mic osimula ion is he name o he echnique known o ep oducing he beha iou o each single
ehicle in he a ic low. This is done by simula ing each ehicle ajec o y wi h a high le el o de ail, his
includes speed, accele a ion, cu en lane, ou e, e c. This means ha mic osimula ion is e y complex,
being many di e en models in ol ed in o he simula ion. As a esul , he e is he need o na ow down
he e iew o he mos c i ical aspec s o sa e y in mic osimula ion. Fo PHOEBE p ojec he ocus will be
on:
1. The ehicle dynamics wi hin he lane such as ca - ollowing beha iou s, bu also be ween lanes
such as lane-changing beha iou s and how sa e/unsa e beha iou is inco po a ed in hese
models,
2. How ulne able oad use s a e cu en ly modelled and conside ed in mic osimula ion ools and
hei in e ac ion wi h o he non-mo o ized and mo o ized a ic a ne wo k le el, and
3. Ne wo k-le el sa e y assessmen app oaches ha a e conside ed/in eg a ed on mic osimula ion
ools (i exis ).
3.1.1 Exis ing mic osimula ion ools
The e a e many so wa e packages ha p o ide o - he-shel solu ions wi h all he equi ed models
in eg a ed in one solu ion. Ba celo (2010) summa izes many a ic simula ion ools a ailable on he
ma ke . The mic osimula ion ones p esen ed a e Aimsun, SUMO, VISSIM, AVENUE, Pa amics,
MITSIMLab and DRACULA. O he ools also exis ha we e no included in Ba celo (2010) as hey we e
no de eloped ye like MATsim.
Wha all hese ools ha e in common is ha hey a e designed in a way ha ehicles always espec
dis ances o o he ehicles and canno e e c ash. This implies ha a ailable a ic simula o ools a e
designed o be sa e by na u e. The ollowing sec ions will s udy mo e in dep h di e en a eas o a ic
simula o ools, emphasizing on how sa e/unsa e beha iou s ha e adi ionally been modelled, and cu en
app oaches owa ds he de elopmen o u u e a ic models ha inco po a e human e o s, so ha
po en ial ehicle c ashes migh happen and sa e y assessmen s migh also be done.
3.1.2 Mo o ehicle mic osimula ion models
3.1.2.1 Longi udinal mo emen s: ca - ollowing models and hei sa e y modeling
implica ions
Ca ollowing (CF) beha iou is one o he main building blocks o mic oscopic simula ion models. I is
desc ibed as a ule (o se o ules) ha desc ibe he longi udinal beha iou o a ehicle on he oad. This
is he accele a ion, speed, posi ion, e c. Some imes is also conside ed as a con ol law, as in
mic osimula ion i con ols wha he ehicles do in he longi udinal dimension, i.e. he di ec ion o a el. I
con ols how he ehicle beha es in he p esence, o absence o ehicles in on .
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
36
These models da e back a long ime when he pionee s in he ield s a ed o p oduce he i s ca ollowing
models. The i s one was de eloped by Pipes (1953). He de eloped a e y simple bu e ec i e ca
ollowing whe e he dis ance be ween ca s inc eases linea ly wi h he speed. A e wa ds Fo bes e al.
(1958), Fo bes (1963) and Fo bes and Simpson (1968) in oduced he eac ion ime wi hin he model, i.e.
how as a d i e can eac o ex e nal changed. This de elopmen was ollowed by Gene al Mo o s
di e en gene a ions o ca ollowing models. These accoun ed by he las gene a ion o eac ion ime,
d i e ’s sensi i i y, ehicles gap and speed di e ence be ween d i e s (Chandle e al., 1958; He man e
al., 1959; Helly, 1959; Gazis e al., 1961).
A b eak h ough was de eloped by Gipps (1981), whe e a mo e complex and ealis ic model was p oposed
ha enabled o be e cap u e d i e s’ beha iou by a se ies o calib a ion pa ame e s. Using he Gipps
(1981) model is possible o emula e bo h a ic s abili y o ins abili ies depending on he calib a ion
pa ame e s. Al hough, o he sake o sa e y analysis his is a model ha is cons uc ed as ‘sa e’ by
de ini ion. This means ha he e canno be ehicle c ashes when using his ca - ollowing.
T eibe e al. (2000) gene alized a model p e iously de eloped o on- and o - amps o wo k wi h gene al
a ic inhomogenei ies making i sui able o mic oscopic a ic simula ion. This is known as IDM
(In elligen D i e Model). A key IDM bene i is ha is de ined as di e en ial equa ions ha can be sol ed
using nume ical me hods. E en wi h low p ecision me hods, he model esul s a e easible bu lack some
o he a ic ea u es ha a ise wi h mo e e ined, and compu a ionally expensi e, me hods o sol e he
equa ions.
Al hough Gipps (1981) and T eibe e al. (2000) and i s de i a i es a e widely used in comme cial so wa e
packages and in esea ch, he models ha e s ong sho comings. Those models can simula e a good
ep esen a ion o a ic lows. Thus, hey p o ide a e age speed, a el imes o lows ha a e close
enough o hose measu ed in eal li e, when p ope ly calib a ed. None heless, hey ha e s ong limi a ions
when i comes o he accu acy o he ehicle’s ajec o ies, his is he accele a ion and speed p o iles o
he ehicles.
This was i s desc ibed by B acks one and McDonald’s (1999), a e wa ds mul iple wo ks ha e done
ex ensi e CF e iews highligh ing hese and mo e d awbacks o he cu en app oaches, especially when
a sa e y analysis wan s o be pe o med (Sai uzzaman and Zheng, 2014; Sai uzzaman e al., 2015; an
Lin and Cal e , 2018; Cal e e al., 2020). ). This is in pa as mos models p esen ed so a a e sa e,
which means ha ehicles ollowing hese con ol laws will ne e c ash. This beha iou is explained by
he ac ha ehicles eac in a pe ec manne o he ex e nal in o ma ion and addi ionally in some models
he ehicles will do wha e e necessa y o a oid colliding wi h o he ehicles. These a e no he only CF
sho comings, as hey ypically omi ha a ehicle can eac o s imulus om ehicles in on o hei
p eceding ehicle. This was in oduced by He man and Ro he y (1965) bu is a imely opic as comme cial
ACC ha e also his ea u e al eady on he oad.
Hamda e al. (2008) and Hamda e al. (2014) de eloped a model o d i e s isk aking beha iou in
ce ain maneu e s such as ea -end collisions du ing ca ollowing scena ios. In addi ion, disc e e e o s
ha e been made o s udy he e ec s o human e o s (e.g. ac ion e o s, cogni i e and decision-making
e o s, obse a ion e o s, in o ma ion e ie al e o s, and iola ions) (S an on and Salmon, 2009) and
dis ac ion (P zybyla e al., 2012) on ca ollowing beha iou and isk. Many ecen s udies such as Be ani
and Chung (2012), Yang and Peng (2010), Paschalidis e al. (2019), Li e al. (2021), Zhang e al. (2022),
Zhang e al. (2023), Ada iko u e al. (2023) p oposed ca ollowing models ha a e capable o aking he
abo e beha iou al unce ain ies in o accoun . Howe e , he e is no e idence o using hese models in he
exis ing comme cial mic osimula ion so wa e packages.
S ill, sa e y analysis can be pe o med using hese sa e CF models. This is done by checking a se ies o
me ics like TTC (Time To Collision), nea mises, ex eme decele a ion alues among he main ones
(McKenna e al., 2006). This way is possible o assess when isky si ua ions happen. None heless, his
leads o some se ious limi a ion, and is ha he CF models p esen ed so a assume a pe ec d i e wi h
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
37
pe ec in o ma ion. This is ha hey do no accoun o any d i e dis ac ion, excessi e isk- aking,
ad e se wea he , pe cep ion mis akes o biases. S an on and Salmon (2009) ound ou ha up o 75% o
oad c ashes ha e human dis ac ion as a con ibu ion. Thus, explici ly modelling some d i e s’ e o s is
meaning ul o a mic osimula ion sa e y analysis.
3.1.2.2 La e al manoeu es: lane-changing models and gap accep ance
Lane-changing models is a decision model ha analyses he need o desi e o a lane change, he bene i s
o a lane change and he easibili y condi ions o a lane change (Ba celó, 2010). The desi e o need o
change depends on he dis ance o he nex u ning and he a ic condi ions. I he d i e is going slowe
han his/he desi ed speed, hen his d i e will y o o e ake he p eceding ehicles. D i e s will also
ha e o check whe he he e will be (o no ) any imp o emen in a ic condi ions a e ha lane change.
D i e s also mus check i he e is enough gap o make he lane change in a sa e way (Ba celó, 2010).
In ecen yea s, he e has been lo s o e o s o e iew and collec all ele an wo k ela ed o lane-
changing models (Toledo e al., 2003; Mo idpou e al., 2010; Rahman e al., 2013; Zheng, 2014; Jian &
Ying-Shuai, 2017; Wang e al., 2019; Ji & Le inson, 2020). Al hough hese models ha e ecei ed less
a en ion compa ed o ca - ollowing models (Zheng, 2014), i is a ma e o g ea in e es o many
esea che s due o i s impo ance in he c ea ion o majo -c ash haza ds. In ac , lane changing is one o
he main causes o oad c ashes on highways (Khel a e al., 2023) and he e o e has nega i ely impac s
on a ic sa e y (Pande & Abdel-A y, 2006; Zheng e al., 2010; Rodeme k e al., 2012).
The i s lane-changing model o mic osimula ion ools was de eloped by Gipps (1986). I conside ed he
necessi y, desi abili y, and sa e y o lane changes. Usually, lane changing beha iou is modelled in wo
s eps: (1) he decision o conside a lane change, (2) he decision o execu e he lance change (Toledo e
al., 2003). Se e al mic oscopic simula o s ha e implemen ed lane changing beha iou s based on Gipps’
model. As an example, Aimsun simula ion ool uses his decision model o ep esen how d i e s change
lanes. Th ee di e en zones a e conside ed: (a) Zone 1, in which decisions a e go e ned by a ic
condi ions o he in ol ed lanes; (b) Zone 2, is he in e media e zone in which d i e s a e looking o a
sa e gap, adap ing hei speed o ind gaps; (c) Zone 3, in which ehicles a e u gen ly eaching hei co ec
lane, educing speed, and looking o ups eam gaps.
Ano he app oach o lane-changing was also p oposed by Spa manns (1978). In his case, he lane-
changing model classi ied lane changes as slowe - o- as e and as e - o-slowe . Depending on d i e s’
speed, i he on ehicles is d i ing a a lowe speed compa ed o he ehicle behind, hen his ehicle
will change o a as e lane o ice e sa. The model is a unc ion o he speed o he on ehicle and he
desi ed speed o he ehicle behind. As an example, Vissim uses a lane-change model which is based on
Spa mann’s model (Ahmed e al., 2021).
Lane-changing models a e ollowed by gap-accep ance models because when a d i e inds a gap o
unde ake he maneu e , he ules go e ning lane-changing models igge immedia ely (Ahmed e al.,
2021). This gap-accep ance model conside s he dis ance o ehicles o a hypo he ical collision poin , hei
speeds and accele a ion a es. And hen he model also calcula es he ime needed by ehicles o
unde ake he ac ion.
Complemen a y o he lane-changing models, he e a e o he models – called look-ahead model – whose
main idea is o p o ide he knowledge o a se o nex u ning mo emen o ehicles in o de o ha e a
clea idea on how o make lane-changing decisions and a oid no eaching he app op ia e u ning lane
(Ba celó, 2010).
Howe e , as obse ed in all hese models, sa e y is conside ed as he key elemen o unde ake
manoeu es. D i e s canno change lane i he e is no a sa e gap be ween ehicles o a sa e dis ance
and speed be ween he p eceding and he ollowing ehicle. As indica ed in he p e ious sec ion ega ding
ca - ollowing models, hese lane-changing models a e also conside ed o be sa e by na u e, which a oid
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
38
d i e s o unde ake decisions ha migh pu a isk o c ashing wi h o he oad use s. The in oduc ion o
human ac o s in hese mic osimula ion models can p o ide a be e ep esen a ion o how d i e s beha e
in he eal wo ld and he e o e, mo e accu a e sa e y analysis migh be able o be pe o med.
3.1.2.3 Human ac o s in ca - ollowing and lane-changing models
To c ea e a mo e ealis ic simula ion model o ehicles esul ing in mo e equen in e ac ions and sa e y-
c i ical con lic s, he e is he need o in oduce human ac o s in he models. This is o d op ou he
un ealis ic assump ion ega ding he d i e capabili ies as s a ed by Boe (1999). In his sense, i is
assumed ha d i e s always aim o hei op imal, achie ing his by applying a unique con ol law, which
migh be he esul o eac ing o s imulus ha a e no able o pe cei e ( o ins ance in bad wea he ). Bu
in eal li e, d i e s ypically adop d i ing s a egies ha , a he han being op imal, a e jus adequa e due
o lack o in o ma ion o ime o check all he al e na i es. Sai uzzaman and Zheng (2014) e iewed he
exis ing (a he ime) s udies on inco po a ion o human ac o s in ca - ollowing models and ound ha
mos o hese models p o ide no psychologically plausible cha ac e iza ion o how humans hink abou
he d i ing ask. This lack o human beha iou elemen s in ca ollowing models is e en mo e acu e when
no ing ha d i ing beha iou is he e ogeneous and can a y depending on indi idual a ibu es such as
age, gende , and so o h.
Thus, he e is a need o imp o e p e iously p esen ed models o inco po a e human ac o s han can
p edic and ep oduce human e o s like longe eac ion imes; while being compu a ionally ac able, and
able o be calib a ed by da a ha can be collec ed. All his while keeping hei abili y o ep oduce gene al
a ic low phenomena as cu en models do.
As an example o isky beha iou s, B acks one e al. (2002) ound ha on he B i ish M27 mo o way 95.8%
o he d i e s was ollowing a headway sho e han he ecommended 2 seconds, and 47.9% o d i e s
ha ing a headway sho e han 1 second. This was no an isola ed case as T eibe e al. (2006) ound
simila esul on a Ge man mo o way, whe e he mos common headways whe e in he ange o 0.9 o 1
second, and some we e as low as 0.3 seconds.
Se e al d i e a ibu es a e co ela ed wi h d i ing s yles. A su ey pe o med ac oss Alabama, US,
d i e s showed ha age and gende a ec s d i e s in his/he pe cep ion o isky si ua ions (Rhodes and
Pi ik, 2011; Ossen and Hoogendoo n, 2011). Also, he su ounding en i onmen in luences he d i e s’
beha iou as ound by Muh e and Voll a h (2011) and T eibe e al. (2006). Sai uzzaman e al. (2014)
made a comp ehensi e lis o mul iple human ac o s om a li e a u e e iew (Hamda , 2012; T eibe and
Kes ing, 2013 among o he s), and hey ound he ollowing ac o s:
• Socio-economic cha ac e is ics (e.g., age, gende , income, educa ion, amily s uc u e)
• Reac ion ime
• Es ima ion e o s: Spacing and speeds can only be es ima ed wi h limi ed accu acy
• Pe cep ion h eshold: Human canno pe cei e small changes in s imuli
• Tempo al an icipa ion: D i e s can p edic a ic si ua ion o he nex ew seconds
• Spa ial an icipa ion: D i e s conside he immedia ely p eceding and u he ehicles ahead
• Con ex sensi i i y: T a ic si ua ion may a ec d i ing s yle
• Impe ec d i ing: Fo he same condi ion d i e s may beha e di e en ly in di e en imes
• Agg essi eness o isk- aking p opensi y
• D i ing skills
• D i ing needs
• Dis ac ion
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
39
• Desi ed speed
• Desi ed spacing
• Desi ed ime headway
To in oduce d i e s’ e o s, especially in ca - ollowing beha iou s, some models ha e been de eloped.
The i s is he one by Wiedemann (1974), who in oduces he d i e s’ pe cep ual h eshold. This e lec s
he inabili y o humans o pe cei e p ecise dis ances and speeds so he e is some ial an e o when a
ehicle app oaches ano he one. This esul s in a d i e ha ing an oscilla ing o e - ime gap wi h he
p eceding ehicle. Figu e 3-1 shows how he ollowing ehicle has oscilla ions in speed and dis ance. o.
Figu e 3-1 Wiedemann’s CF model (Sou ce: Wiedemann, 1974).
Fellendo and Vo isch (2010) implemen ed a modi ica ion o he Wiedemann model in Vissim
mic osimula ion ool. The p oblem is ha no clea calib a ion me hodology has been es ablished, al hough
se e al a emp s ha e been p o ided in he li e a u e (Pa k and Qi, 2006; Gomes e al., 2004; Iš oka
O ko ic´ e al., 2013; Lownes and Machemehl, 2006). This is one o he key d awbacks on in oducing
human ac o s in o hese mo o ehicle simula ion models. The calib a ion and alida ion a e needed and
he ypical da a a ailable is no good enough o ha pu pose.
So a , he p esen ed models enable o cha ac e ize he pe cep ion e o s, bu his does no accoun he
isk- aking beha iou o some d i e s. Based on he p ospec heo y o Kahneman and T e sky (1979),
which is accep ed as a human decision-making model o ou comes ha can be isky, Hamda e al. (2008)
and Hamda e al. (2014) de elop a d i ing model based on he ac ha d i ing is he sequence o
po en ial isk- aking decisions. This is achie ed by cha ac e izing he d i e s wi h some biases o some
po en ial ou comes. In his case, a d i e ’s subjec i e p obabili y o being in ol ed in o a ea end collision
is also included.
None heless, he concep o ‘human e o ’ emains o be cha ac e ised. This is a e y wide de ini ion ha
adi ionally ha e been classi ied as “in olun a y e o s and iola ions” (Reason, 1990). A iola ion is a
delibe a e ac o he d i ing disobeying egula ions, like speeding. An e o is any beha iou ha di e s
d i e s om he desi ed d i ing happening un olun a y. This can be u he subdi ided in o i e di e en
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
40
ca ego ies: ac ion e o s, cogni i e and decision-making e o s, in o ma ion e ie al e o s, obse a ion
e o s, and iola ions (S an on and Salmon, 2009).
In he pa icula case o ca - ollowing models, human ac o s can be inco po a ed in o hese exis ing
models. Be ani and Chung (2012) imp o ed Gipps’ (1981) model by conside ing human impe ec ion in
p ocessing in o ma ion and execu ing ac ions. Mo e speci ically, hey included human pe cep ion
limi a ions in de ec ing speed di e ences, ex a delay in d i ing phase changes – assuming ha eac ion
ime inc eases a e being in a ixed si ua ion; ha is, ei he in a cons an speed o in an accele a ion
phase – and d i e impe ec ion in adjus ing speeds. Howe e , human e o s, such as dis ac ion and isk
aking, a e omi ed in hei model. Hamda and Mahmassani (2008) did also explo e he modi ica ion o
exi ing ca - ollowing models o accoun o some human ac o s, achie ing some ehicula c ashes. In a
simila way, T eibe e al. (2006) pu posed ex ensions o he IDM model so i could ha e es ima ion e o s,
ini e eac ion imes, spa ial an icipa ion and empo al an icipa ion. They a gued ha he model sol ed he
un ealis ic ehicula dynamics and c ashes ha p e ious human- ac o models had no .
The la es ele an e o in his ield is Van Lin and Cal e (2018) and Cal e e al. (2020). In Van Lin
and Cal e (2018), a me hodological amewo k o design a ca - ollowing model conside ing human
ac o s is p esen ed. This wo k ackled on which human ac o needs o be modelled and in which
hie a chy. As hey p oposed, he e a e se e al cogni i e le els aking decisions when d i ing. As shown
a Figu e 3-2, he e is he s a egic le el ha con ols he ou e, he manoeu ing le el ha con ols he
lane and inally he con ol le el, which is he ca - ollowing le el.
Figu e 3-2 The hie a chical s uc u e o he oad use ask. Pe o mance is s uc u ed a h ee in e wined le els.
(Sou ce: an Lin and Cal e , 2018).
Ano he esea ch s eam abou oad use beha iou in a ic mic osimula ion has ocused on such
beha iou in lane changing maneu e s. Ulak e al. (2019) in es iga ed he e ec s o age and expe ience
o d i e s on lane changing beha iou o d i e s and ound ha a ied isk pe cep ion o d i e s depending
on hei age and expe ience a ec d i ing beha iou and ul ima ely in luence a ic sa e y and
pe o mance measu es. Das and Ahmed (2022) s udied he d i ing beha iou unde ad e se wea he
condi ions (i.e. poo isibili y) in a ic mic osimula ion and ound ha adjus ed a ic mic osimula ion
pa ame e s in ad e se wea he condi ions p oduce lowe a e age speeds wi h highe o al a el imes
and o al delays han clea wea he .
Wi h he in oduc ion o au oma ed ehicles and he ambi ion o adop hese ehicles in he nea u u e, a
ew s udies ha e ocused on he e ec s o human beha iou on d i ing in mixed a ic (i.e. a ic wi h
con en ional and au oma ed ehicles). De Zwa e al. (2023) s udied he e ec s o d i ing adap a ion
beha iou in e ms o ime headway, ollowing dis ance, and eloci y o d i e s in he p esence and
absence o au oma ed ehicles and ound ha d i e s adap hei headway and speed in such
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
47
models is ha i is concei ed o be “disc e e in na u e” (Jas i & Higgs, 2006), meaning ha i does no
model he con inuous in e ac ion o mixed a ic low (Qu e al., 2017). In ac , depending on he cellula
space di ision, he simula ion will be mo e/less accu a e. As opposed o his, social o ce models simula e
pedes ian lows as a con inuous model. In his sense, indi iduals can change he di ec ion and magni ude
o hei speeds acco ding o he o ces ha a e applied o each o hem, as indica ed on New on’s second
law o dynamics. Ano he app oach o ealis ic modeling o mo emen s and in e ac ions in oad simula ion
en i onmen s is he agen -based modeling (Mohammed e al, 2019). ABM o e s he possibili y o
cap u ing he complexi y o eal-wo ld beha iou (Bas e e al., 2013; Hussein & Sayed, 2017). Recen ly,
one s udy used a da a-d i en app oach i.e. In e se Rein o cemen Lea ning (IRL) algo i hm, o simula e
and in es iga e cyclis s in e ac ion wi h pedes ians (Alsaleh and Sayed, 2022) and ound ha hese
models can accu a ely p edic he ajec o ies o cyclis s.
Exis ing a ic mic osimula ion ools, such as Aimsun, Vissim, Sumo, u ilize wo models o independen ly
model he mo ion o oad mo o ized use s: ca - ollowing and lane-changing models ha simula e
longi udinal and la e al mo emen s (Tawddle e al., 2014). Bu o cyclis s, his beha iou seems no eal
and he e o e i is no enough o use hese ca -based models o mimic cyclis s’ beha iou . Ald ed e al.
(2019) s udied he impac o cyclis s in sha ed bus lanes in he cen e o London. The au ho s used Vissim
mic osimula ion ool o simula e bicycles and buses sha ing he same lane, bu wi h a manipula ion o
some model pa ame e s o be e ep esen – app oxima ely – he beha iou o cyclis s. They also
indica ed ha he ypical app oaches o mic osimula ion modeling so wa e ools a e p oblema ic o
ep esen ing cyclis s.
In he case o Aimsun simula ion so wa e, since he elease o Aimsun Nex e sion 20, bicycles can be
modeled by using a “non-lane-based beha iou ” which be e ep esen how cyclis s and mo o bike ide s
beha e (Hayman, 2020). In his case, mul iple ehicles can occupy he same longi udinal posi ion wi hin
a lane as shown in Figu e 3-4, and his allows cyclis s o o e ake o he ehicles and o m g oups o
cyclis s. The model is also capable o ins uc ing ehicles o s ay close o he le / igh side o he oad
depending on he di ec ion o a el. A ecen in es iga ion o he e ec s o spo cyclis s on ehicula
a ic on na ow wo-lane u al oads modelled hem in mic osimula ion and ound ha bicycle a ic
dec eased a e age ehicula speed and inc eased ehicula a ic delays (Moll e al., 2021).
Figu e 3-4. Sc eensho s o bicycle- ehicles in e ac ion in Aimsun Nex . On he le , he adi ional modeling app oach
o bicycles conside ed as ehicles. On he igh , he new implemen ed non-lane-based beha iou model in Aimsun
Nex . Sou ce: Adap ed om Hayman (2020).
I is clea ha mic oscopic simula o s can be con igu ed o simula e mixed a ic condi ions (e.g. be ween
pedes ians and cyclis s), by al e ing some pa ame e s and wi hou changing exis ing beha iou al models.
Howe e , he esul s ha he models will p o ide may be un ealis ic. The e o e, he e is a need o
implemen ing new beha iou al models o simula e a ic on pa hs o spaces ha a e sha ed among use s.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
48
Despi e i s cu en impo ance, li le a en ion has been paid o model he in e ac ion o wo-wheele and
o he oad use s (Sh i as & Kuma , 2023). The way wo-wheele d i e s mo e is di e en om a ca and
he e o e u he models need o be examined. In his sense, Sh i as & Kuma (2023) p oposed an
ex ension o he SFM o inco po a e in e ac ions be ween pedes ians, ca s, and wo-wheele s a a
signalized in e sec ion. The au ho s inco po a ed a d i ing o ce e m in he desi ed di ec ion o mo emen ,
a epulsi e e m o a oid any physical con ac and main ain a sa e dis ance and a ca - ollowing ea u e
and con lic a oidance e ms in a ma hema ical model. Qu e al. (2017) modi ied and ex ended he SFM
o mixed a ic low conside ing he in e ac ion be ween elec ic bikes and ca s. Thei ex ended model
also included lane-changing beha iou o e-bikes and con lic a oidance beha iou . No e ha elec ic
bikes ha e highe speed han non-elec ic bikes, and his means elec ic bike ide s end o change lane
mo e equen ly and p e e o change lane in o he ehicles lane when i is no occupied. Meh a e al.
(2019) analyzed he o e aking beha iou o mo o ized ehicles when passing bicycles on u ban oads.
In es iga ing he beha iou o cyclis s h ough mic osimula ion has been he ocus o a ew p e ious
ho izon p ojec s oo. Fo example, he ALLEGRO4 p ojec exploi ed a wide ange o da a collec ion
me hodologies such as Wi-Fi and 3D senso s, and ex ensi e su eys and c ea ed a mic osimula ion
model o cyclis s in ope a ional le el, mid- e m ou e choice decision making, and long e m ou e choice
decision making.
3.1.5 E-scoo e ide beha iou and in e ac ions wi h o he use s
Gi en new o ms o e-mobili y ha e only eme ged in ecen yea s, esea che s ha e only jus s a ed
looking in o he beha iou o new o ms o mic omobili y oad use s such e-scoo e s. Mos o hese s udies
ha e been conduc ed in an empi ical-simula ion ashion, meaning ha hey ha e i s collec ed empi ical
da a on e-scoo e s a ic pa e ns and ha e hen c ea ed a simula ion en i onmen o s udy hei
beha iou .
• Mclean (2021) and Mclean e al., (2021) collec ed empi ical da a on e-scoo e pilo p og am
in he Ci y o Calga y in Canada and c ea ed a syn he ic wo kload model o e-scoo e a ic.
They s udied he impac o e-scoo e s and hei managemen policies (such as lee size,
ba e y e-cha ging s a egies, and u ban pa king in as uc u e loca ions) on a el demand
and cos and highligh ed he impo ance o p ope si e selec ion o pa king a eas and ba e y
cha ging in as uc u e.
• La inopoulos e al., (2021) in es iga ed he impac o e-scoo e s on a ic wi h su eys, expe
in e iews and mic osimula ion o Pa is and ound ha p ope adop ion o his mic omobili y
mode o anspo needs a holis ic analysis o beha iou al and cul u al elemen s. They u he
ecommended ha any ci y conside ing e-scoo e s as pa o hei mobili y o e ing should
conside he esul s in he cul u al and policy con ex o he ci y o gene a e a se o guidelines.
• Pazzini e al., (2022) in es iga ed he a ic beha iou o e-scoo e s in he ci y o T ondheim
in No way and ound ha a ound 60% o 90% o e-scoo e s used bicycle pa hs depending on
he oad en i onmen con ex .
• Finally, Kazemzadeh and Sp ei (2022) conduc ed a sys ema ic e iew o e-scoo e s udies a
mic oscopic le el and ound ha hei is a ely conside ed in he li e a u e as a dis inc g oup
o oad use s. They sugges ha u u e s udies should analyse he in e ac ion o e-scoo e s
wi h o he oad use s, pa icula ly pedes ians.
4h ps://doi.o g/10.3030/669792
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
49
3.1.6 Mic osimula ion models o mixed a ic (au onomous and con en ional
ehicles)
The inc easing p esence o connec ed and au onomous ehicles (CAVs) on anspo ne wo ks has g own
in e es in analyzing he in e ac ion be ween au onomous ehicles and o he ulne able oad use s (Ta idis
e al., 2019). The ole o CAVs in imp o ing a ic sa e y is clea : less human-dependen ehicles and
hence ewe human e o s, which is ansla ed in o less c ashes (Singh, 2015) – al hough new ype o
c ashes migh also appea . As eal-wo ld da a is no a ailable ye , mic oscopic simula ion me hods p o ide
an al e na i e app oach o es how a ic sa e y could change unde di e en pene a ion a es o CAVs
(Ta idis e al., 2019; Tumminello e al., 2023). This in oduc ion o CAVs in o he a ic ne wo k will be a
g adual p ocess and he e o e, he e will be a ansi ion pe iod in which CAVs and human-d i en ehicles
will coexis (Ren e al., 2023) – also known as mixed a ic en i onmen .
Papadoulis e al. (2019) s udied he impac o CAVs unde di e en pene a ion a es on mo o ways. The
au ho s de eloped a decision-making CAV con ol algo i hm in Vissim o allow a CAV o ha e longi udinal
con ol, sea ch adjacen ehicles, iden i y nea by CAVs and make la e al decisions. The sa e y impac
assessmen was conduc ed using he SSAM model. Resul s showed ha CAVs educed a ic con lic s
e en a low pene a ion a es. Elawady e al. (2022) also analyzed he impac o CAVs on a ic sa e y in
an in e sec ion using SSAM app oach and unde di e en pene a ion a es, p o iding simila esul s.
Tumminello e al. (2023) also in es iga ed he impac o CAVs on he ope a ional pe o mances o a single-
lane oundabou using Aimsun Nex .
Howe e , he implica ions o au onomous ehicles on ulne able oad use s such as pedes ians,
bicyclis s, scoo e s a e s ill no clea (Ku ela e al., 2022). In ac , he in e ac ion be ween CAVs and VRU
in a mixed a ic en i onmen has no been well s udied. Al hough CAVs a e p og ammed o gi e p io i y
o VRUs, i is s ill unclea how hese use s will eac o hese ehicles. Ku ela e al. (2022) in es iga ed
he in e ac ion o VRU on CAVs by explo ing he in ol emen pa e ns o VRU in CAVs c ashes. Resul s
showed ha VRU we e a aul in a la ge pe cen age o c ashes ha di ec ly in ol ed VRUs. The au ho s
men ioned ha u u e s udies could also explo e his opic wi h a la ge da ase due o he cu en CAV-
VRU collision da a limi a ion. Kala ian & Fa ooq (2021) p oposed a machine lea ning amewo k o explo e
ac o s a ec ing pedes ians’ wai ing ime be o e c ossing in he p esence o au oma ed ehicles. The
au ho s de eloped a i ual eali y expe imen o collec beha iou al da a. Resul s showed ha he
p esence o au oma ed ehicles, among o he s, we e he main con ibu ing ac o s o longe wai imes.
Fewe s udies ha e been dedica ed o analyze he in e ac ion be ween bicyclis s and CAVs.
Despi e he inc easing numbe o s udies ha analyzes he impac o CAVs on a ic sa e y and ope a ion,
he li e a u e on he sa e y implica ion o CAVs on VRU is s ill limi ed and needs u he esea ch o ge a
clea pic u e o hei po en ial impac .
3.1.7 Using mic osimula ion models o assess sa e y
P e ious sec ions desc ibed he cu en s a e-o - he-a in mic oscopic modelling app oaches and hei
implica ions in sa e y analysis. Fo ins ance, how mo o ized ehicles o non-mo o ized use s mo e on a
sec ion le el, based on which ules and how hey in e ac wi h he su ounding ehicles, adap ing hei
dis ance and speed, and changing lanes o c ossing in e sec ions i a sa e gap exis s. Fu u e po en ial
imp o emen s o hese models will ocus on be e ep esen ing use s’ beha iou s on mic osimula ion
models o allow mo e ‘unsa e’ beha iou s. This in ac migh be achie ed when ‘human e o s’ a e
inco po a ing in he models, which may in oduce he concep o unsa e y in d i e s’ beha iou .
Howe e , all hese app oaches a e mainly ocused a a mic oscopic le el. His o ically, he sa e y
assessmen o oad sec ions (and as a esul , oad ne wo ks) has been done based on his o ical c ash
eco ds, s a is ical modelling and/o p o essional ield obse a ion (Mahmud e al., 2019). Howe e , hese
app oaches migh no be eliable due o c ash epo ing and biased judgemen . In addi ion, hese a e also
called eac i e app oaches as hey need a conside able amoun o p e ious c ash da a o e alua e sa e y.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
50
Because o hese limi a ions, esea che s explo ed new p oac i e me hodologies o e alua e a ic sa e y.
These app oaches e alua e a ic sa e y in mic oscopic a ic simula ion models based on he
compu a ion o sa e y me ics by using simula ion ou pu s. The ad an age o his app oach is ha he e is
no need o al e d i e s’ beha iou o make he simula ion unsa e and allow collisions, which in pa
simpli ies he p oblem. Also, no eal c ash da a is equi ed o make an ini ial es ima ion o oad sa e y.
Such sa e y measu es a e adi ionally called Su oga e Sa e y Measu es (SSM). These app oaches ied
o iden i y speci ic su oga e indica o s ha a e highly connec ed o he isk o collisions, which occu s
mo e equen ly han c ashes. The usage o mic osimula ion modelling has imp o ed he eliabili y o his
su oga e app oach, as human obse e s a e no longe needed o da a collec ion (Mahmud e al., 2019).
O e he pas decades, he e has been lo s o e o s o de elop su oga e sa e y measu es. To measu e
he closeness o a collision, he mos used su oga e measu es a e: he ime o collision and pos
enc oachmen ime (Zheng, e al., 2021). Time o collision (TTC) is de ined as he ime emaining p io o
a collision when he speed and ajec o y o ehicles a e main ained as cons an (Haywa d, 1971). Simila
indica o s o TTC include ime o collision (Hydén, 1987), ime-exposed TTC (Minde houd & Bo y, 2001),
ime-in eg a ed TTC (Minde houd & Bo y, 2001) and modi ied TTC (Ozbay e al., 2008). Pos
enc oachmen ime (PET) is he ime di e ence be ween a ehicle lea ing he a ea o enc oachmen and
a con lic ing ehicle en e ing he same a ea (Coope , 1984). O he simila indica o s include enc oachmen
ime (Allen e al., 1978), gap ime (Ge man & Head, 2003), and headway (Vogel, 2003). Bo h TTC and
PET ep esen a measu e o empo al p oximi y, in he case o TTC ep esen s he p oximi y o a po en ial
collision poin on he p esence o a collision cou se, whe eas PET ep esen s he ac ual sa e y ma gin
wi hou any equi emen o a collision cou se (Zheng & Sayed, 2019). The Fede al Highway
Adminis a ion o he US Depa men o T anspo a ion de eloped a ool o assess su oga e sa e y
measu es called SSAM (Pu e al, 2008), which will be desc ibed mo e in de ail in he ollowing sec ion.
O he indica o s, such as decele a ion and angle-based indica o s, ha e been p oposed in he li e a u e
o measu e he in ensi y o e asi e ac ions o a oid collision – such as b aking o swe ing. The mos
common and used decele a ion-based indica o is he decele a ion a e o a oid a c ash (Coope &
Fe guson, 1976), and o he simila indica o s such as decele a ion o sa e y ime (Hup e , 1997).
Despi e he usage o hese su oga e sa e y measu es in simula ion, none o hese a e sys ema ically
selec ed o locally alida ed. The e ha e been li le o no a emp s o in eg a e ne wo k-le el oad sa e y
assessmen me hodologies in mic osimula ion modeling ools. To he bes o he e iewe s’ knowledge,
he SSAM app oach is he only oad sa e y e alua ion echnique ha has ied o in eg a e mic osimula ion
ools and su oga e sa e y measu es o assess sa e y impac s on a ne wo k le el. The iRAP S a Sco e is
ano he wo ld-wide used app oach o e alua e sa e y in a s a ic way o u ban oad ne wo ks. The ollowing
sub sec ions desc ibe bo h app oaches, emphasizing he ad an ages and disad an ages o hei
applica ion and in eg a ion wi h mic osimula ion ools.
3.1.7.1 Vehicle ajec o y simula ion o sa e y assessmen : he SSAM app oach
An app oach ha has been adi ionally u ilized o e alua e sa e y is based on pos -p ocessing echniques
ha use da a ou pu om a ic simula o s. Compa ed o ield-based obse a ions, simula ion ools a e
able o collec da a in a mo e quick and accu a e way (Ge man & Head, 2003; Cun o, 2008). Recen ly,
simula ion ools a e he mos used app oaches o sa e y s udies as hey ha e he capabili y o
au oma ically collec simula ed a ic da a and analyze po en ial a ic con lic s (Wang & S ama iadis,
2013). In his case, one o he mos used models o iden i y a ic con lic s is he Su oga e Sa e y
Assessmen Model (SSAM) (Pu e al, 2008). SSAM is a ool, de eloped by he Fede al Highway
Adminis a ion o he US Depa men o T anspo a ion, ha uses ehicle ajec o y da a o iden i y a ic
con lic s. This ajec o y da a is ob ained om mic oscopic a ic simula ion models and is used o obse e
ehicle- o- ehicle in e ac ions. In doing so, his app oach is able o iden i y a eas ha a e mo e p one o
inciden s and he e o e, assess sa e y impac on a ne wo k le el. SSAM algo i hms p ocess one simula ion
ime s ep a a ime and checks o a ic con lic s using he p ede ined sa e y Time o Collision and Pos
Enc oachmen Time h eshold alues as su oga e sa e y measu es. In his sense, SSAM simula es u u e
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
51
posi ions o ehicles ollowing hei ajec o y and keeping he same speed o he nex simula ion s eps
and up o he du a ion o he p ede ined TTC (de aul alue is 1.5 s). I he e is an o e lap o ehicles, a
con lic will be iden i ied. The la es e sions o SSAM also p o ide hea maps showing he concen a ion
o con lic s along he simula ed oad ne wo k. Figu e 3-5 p esen s he p ocess o in eg a e a ic simula ion
models and SSAM.
Figu e 3-5 In eg a ion o a ic simula ion models and SSAM
Recen ly, he SSAM app oach has been ex ensi ely used in sa e y e alua ion and i is e en ecognized
as he only easible me hodology o use mic oscopic simula ion o a ic sa e y assessmen (Sha,2020).
Caliendo & Guida (2012) ied o iden i y a ela ionship be ween simula ed con lic s and eco ded c ashes
on unsignalized u ban in e sec ions in I aly. The au ho s wan ed o p o e ha mic osimula ion models
we e a po en ial ool o assessing sa e y. The me hodology he au ho s p esen ed was he use o
mic osimula ion – in his case, Aimsun a ic simula ion so wa e – o model a ic low du ing peak hou s.
And he SSAM was also used o calcula e and classi y con lic s. Resul s showed a signi ican co ela ion
be ween eco ded c ashes and c i ical con lic s ob ained om mic osimula ion models. The e o e, he
au ho s admi ed ha he app oach was a alid al e na i e o p edic c ashes, despi e some iden i ied
limi a ions such as he unmodelled “illegal” o m o d i e s o he quan i ica ion o c ash se e i y in he
analysis.
Hab emichael e al. (2014) e alua ed sa e y implica ions o agg essi e d i ing using a mic oscopic a ic
simula o app oach. In his case, he au ho s used SSAM in combina ion wi h VISSIM o analyze he c ash
isk, se e i y le el and he pe cei ed bene i s o agg essi e d i ing unde conges ed and non-conges ed
condi ions. In his sense, some p opo ions o ehicles we e simula ed o beha e agg essi ely by
modi ying some d i e beha iou pa ame e s o ca - ollowing and lane-changing models in VISSIM. Th ee
ways o oad agg essions we e conside ed: (1) speeding abo e he legal speed limi o d i ing oo as o
he a ic condi ions; (2) ollowing closely; (3) wea ing h ough a ic by ab up change o lanes. The da a
ou pu om he simula ion – in his case, ajec o ies o ehicles – was used as an inpu in o he SSAM
model o de ec a ic con lic s.
Wang e al. (2018) e alua ed in e sec ion sa e y h ough a combina ion o mic osimula ion and Ex eme
Value Theo y (EVT). They simula ed 10 u ban in e sec ions in VISSIM and used ield da a o calib a e he
model. Vehicles ajec o ies om he simula ion model was used as an inpu in o he SSAM which was
used o ex ac con lic da a. Ex eme Value Theo y (EVT) based me hods we e used o model
simula ed/ iled con lic da a and de i e he Es ima ed Annual C ash F equency (EACF), used as Su oga e
Sa e y Measu es (SSM). The au ho s conclude ha his combina ion o mic osimula ion and EVT was a
p omising choice o sa e y e alua ion.
As a i a e al. (2019) compa e di e en a ic simula ion models (Aimsun, VISSIM and T i one) o e alua e
sa e y o ypical in e sec ions ( oundabou , a ic ligh egula ed in e sec ion and an un egula ed
in e sec ion). In o de o ep oduce a ic sa e y measu es, he sa e y e alua ions a e ca ied ou by using
SSAM, which e alua es he su oga e sa e y measu es by using ehicle ajec o ies. The au ho s used
he TTC, PET, Maximum Speed (MaxS) and Decele a ion Ra e (DR) as a su oga e sa e y measu es. The
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
52
au ho s men ioned ha esul s we e simila be ween di e en mic oscopic models. They concluded ha
SSAM analysis showed a g ea e dange in he scena io wi h he oundabou (lowe TTC). All h ee
simula o s showed how a signal egula ed in e sec ion was much sa e han a oundabou and sa e han
he un egula ed in e sec ion.
Fu he imp o emen s on he p e ious wo k we e de eloped by As a i a & Gio é (2019). The au ho s
in oduced a way o e alua ing sa e y based on inco po a ing d i e e o s and iden i ying he esul ing
c ash consequence using SSAM. The p oposed me hodology has been applied in VISSIM, Aimsun and
T i one. The au ho s p oposed o simula e a d i e dis ac ion by examining all ehicle ajec o ies e e y
second o he simula ion and pe o m a “dis ac ion” on he ajec o y on ce ain ehicles a a cons an
speed. I he e is a po en ial collision, he ex en o he collision is e alua ed. Th ee main pa ame e s a e
conside ed in hei implemen a ion: he simula ion ime s ep, he dis ac ion ime, and he angle o he
possible ehicle ajec o y. Wi h his app oach, wo possible esul s can be possible: (1) he ehicle is
in ol ed in a c ash, which is de ec ed by e alua ing he poin s o in e sec ion be ween wo objec s, and
(2) he ehicle does no impac wi h any o he one. Howe e , wi h his app oach, he e is a need o a
de ailed calib a ion o ce ain pa ame e s, such as he a e o dis ac ion, which is a p io i unknown.
This ajec o y-based app oach has also been used o s udy he sa e y impac o connec ed and
au onomous ehicles (CAVs) using a mic oscopic s udy. In pa icula , Miqdady e al. (2021) applied a
simula ion-based sa e y assessmen o e alua e he numbe o inciden s a e he in oduc ion o
connec ed and au oma ed ehicles. The modelling o CAVs was done using Aimsun Nex API. The
po en ial con lic s we e ob ained by using SSAM and calcula ing he TTC indica o . The au ho s used a
TTC h eshold (1.5s) a e unde aking a sensi i i y analysis and esul s showed a signi ican di e en
be ween ex eme alues (0.5 and 2.5s). Resul s demons a ed ha he highe he pene a ion a e o
CAVs, he ewe numbe o po en ial c ashes.
Ano he example is he Ho izon2020 unded LEVITATE p ojec (2021), which e alua ed he sa e y
impac s o CAVs ollowing he same ajec o y-based app oach, he SAFE-UP p ojec (2021) did a simila
wo k also including some assessmen s wi h he VRUs in e ac ions. Sha e al. (2022) s udied he sa e y
impac s o dedica ed lanes o CAVs using mic osimula ion. The me hodology in eg a ed Aimsun as a
a ic simula o wi h SSAM o iden i y a ic con lic s h ough ehicula ajec o ies. In his case, he
au ho s p esen ed a new app oach o es ima e he expec ed numbe o c ashes by using a ic con lic
da a ob ained om mic osimula ion. This was ca ied ou by using a p obabilis ic me hodology p oposed
by Ta ko (2018). The i-DREAMS5 Ho izon 2020 p oposed di e en de ini ions o oad isk on he basis o
mic oscopic na u alis ic d i ing da a collec ed in 5 coun ies, i s indica o s, and i s con ibu ing ac o s om
a mic o sa e y assessmen pe spec i e.
Recen ly, a new pape de eloped by Leona di & Dis e ano (2023) compa ed he ope a ional and sa e y
pe o mance o mul i-lane oundabou s and u bo- oundabou s. The au ho s ollowed he same app oach
as he p e ious wo k: hey simula ed a ic mo emen s using a mic oscopic simula o and hen used he
ehicle ajec o y da a ou pu as an inpu o he SSAM o e alua e sa e y impac s. The au ho s also
men ioned ha hey p e e ed o choose Aimsun so wa e because i was mo e eliable han VISSIM in
gene a ing mo e ealis ic ajec o ies, which is ex emely impo an i SSAM is used a e wa ds.
O he s udies ha e also e alua ed he impac s o oadside sa e y imp o emen s by using a compu e
so wa e called Roadside Sa e y Analysis P og am (RSAP) (Mak e al., 1998), which was de eloped as a
ool o he 2011 Roadside Design Guide (1996), published by he Ame ican Associa ion o S a e Highway
and T anspo a ion O icials (AASHTO). This ool only conside s isola ed d i e s c ashes agains ba ie s
o oadside obs acle wi hou conside ing he in e ac ion be ween ehicles. The so wa e pe o ms a se ies
o simula ions o enc oachmen s o a ypical oadway segmen and es ima es he c ash cos s.
5 h ps://id eamsp ojec .eu/wp/
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
53
The ollowing Table 3.4 shows a summa y o all e iewed models ha e alua e a ic sa e y by analyzing
he ehicle ajec o y ob ained om mic osimula ion models.
T a ic simula ion me hodology has been widely used and become a common p ac ice in anspo a ion
enginee ing by o e ing he abili y o analysing complex anspo a ion aspec s. In e ms o sa e y, he use
o mic oscopic simula ion (o mic osimula ion) is becoming p og essi ely easible as well, as se e al
app oaches using sui able me hodological amewo ks and ools quan i ying su oga e sa e y measu es
become eadily a ailable. Fo his case, he mos common echnique is he con lic -based app oach ha
enable anspo a ion enginee s o in es iga e high- isk loca ions wi hou he need o ield c ash da a
(Ghanim e al., 2020).
One o he mos common ways o es ima e con lic s using mic oscopic models is he Su oga e Sa e y
Assessmen Model (SSAM) so wa e, a model de eloped by Fede al Highway Adminis a ion (Pu &
Joshi, 2008). The so wa e analyses he ehicle ajec o y da a and iden i ies con lic s. Speci ically, a
con lic is iden i ied when he Time-To-Collision (TTC) and Pos -Enc oachmen Time (PET) a e lowe om
p ese h esholds, as iden i ied in ea ly s udies explo ing he possibili y o using mic oscopic simula ion o
oad sa e y assessmen s (As a i a e al., 2011). In addi ion, he iden i ied con lic s a e also classi ied in o
h ee manoeu e ypes, namely ea -end, lane change and c ossing con lic s based on con lic angle. A
a ie y o mic osimula ion s udies ha e been iden i ied con lic s using he SSAM o e alua e
consequences o di e en anspo a ion planning (Goh e al., 2014; P es on and Pulugu ha, 2021),
con ol policies (Ribei o e al. 2019; Li and Sun, 2019; K onp ase e al., 2020; Shahdah and Azam, 2021),
oad con igu a ions (Giu è e al., 2019; Ghanim e al., 2020; Bahmankhah e al., 2022) as well as
anspo a ion inno a ions (Xin e al., 2019; Mou akos e al., 2021; Elawady e al., 2022) in e ms o a ic
sa e y.
As men ioned, he mic oscopic simula ion has been ex ensi ely used wi h ega ds o he su oga e sa e y
measu es e alua ion. Fo his eason, se e al s udies ha e a emp ed o examine he alidi y o he esul s
de i ed om mic oscopic models. Findings om a simula ion s udy indica ed a s ong ela ionship
be ween su oga e sa e y measu es o a ic simula ion and eal c ash da a (Ozbay e al., 2008).
Speci ically, he accu acy o con lic p edic ion was e ealed when compa ing simula ed con lic s wi h ield
con lic s (Zheng e al., 2019; Essa and Sayed, 2020). In a di e en s udy, So e al. (2015) p oposed an
app oach o in eg a ing ehicle dynamics models wi h a ic simula ion models aiming in gene a ing
ealis ic ehicle ajec o ies, and compa ed i s esul s wi h adi ional a ic simula ion model ones p o ing
ha he p oposed app oach in eg a e a s onge co ela ion wi h eal c ashes. Finally, con lic s de i ed
om mic oscopic simula ion models we e also compa ed o p edic ed con lic s using s a is ical modeling,
and i was e ealed ha he simula ed con lic -based me hod indica ed a signi ican co ela ion in he
esul ing ou comes and had a be e pe o mance in iden i ying high- isk loca ions (So e al., 2015).
As a i a e al. (2020a) p oposed a no el su oga e sa e y indica o ounded on ehicle ajec o ies, while
simul aneously conside ing oadside objec s. The heo e ical model is alida ed h ough compa isons
be ween he calcula ion o su oga e sa e y measu es on ajec o ies ob ained om mic osimula ion and
c ash isk ob ained h ough eal c ash da a ob ained by se e al u ban in e sec ion scena ios. Apa om
con lic gene a ion, when he ajec o ies we e o e lapping (o close enough), TTC and PET indica o s
we e also examined, while he au ho s concluded ha a model including mean ene gy is s a is ically
equi alen o a collision-based model.
One majo limi a ion o applying a ic simula ion in e alua ing sa e y is he absence o comple e and
es ablished models o simula ing po en ial c ashes. A ew s udies ha e a emp ed o o e come his
issue h ough p oposing al e na i e me hodologies. In pa icula , Guido e al. (2019) p esen ed some
applica ions o a new p ocedu e based on po en ial c ash e en s o e alua ing o sa e y le els in
mic oscopic simula ion models aking in o accoun also po en ial c ashes wi h oadside objec s and
ba ie s. In addi ion, a simila s udy conduc ed by Hab emichael and De Picado San os (2014) used
simula ed con lic s o e alua e odds a io o c ash in ol emen . Mo eo e , Shahdah e al. (2014) used
simula ed con lic s linked o mally o obse ed c ashes o in es iga e an al e na i e app oach o
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
54
es ima ing C ash Modi ica ion Fac o s (CMFs). Fu he mo e, a ecen s udy by Oikonomou e al. (2023)
p oposed a me hodological amewo k o es ima ing c ash a es along wi h ne wo k cha ac e is ics and
a ic measu es using a mic osimula ion con lic -based analysis. Finally, a alida ion o es ima ed c ashes
om simula ed con lic s showed ha c ashes we e unde es ima ed, ne e heless mo e accu a e and
eliable p edic ed c ashes we e de i ed wi h a p ope simula ion ne wo k calib a ion (Zheng e al., 2019).
3.1.8 Risk assessmen o oad geome ic con igu a ion
Focusing on oad design, usually he e is a lack o na u alis ic d i ing da a, esul ing in no eliable oad
geome ic designs implemen a ion wi h limi ed ope a ional and sa e y pe o mances. In addi ion, he
iden i ica ion o he op imum geome ic design equi es de ailed obse a ions and e alua ions o di e en
geome ic con igu a ions. The a ic mic osimula ion me hod is conside ed as an ideal me hod o his kind
o in es iga ions as he e is he abili y o es di e en con igu a ion scena ios and e alua e bo h sa e y
and pe o mance. The e o e, he mic oscopic simula ion me hod has been ex ensi ely used o e alua e
consequences o di e en a ic planning and con ol policies. Fo ins ance, se e al ecen s udies
in es iga ed a ic ope a ions a oundabou s (Giu è e al., 2017; Giu è e al., 2018; Giu è e al., 2019;
Ghanim e al., 2020; Bahmankhah e al., 2022) by es ing di e en con igu a ions. Fu he mo e, ano he
mic osimula ion s udy examined impac s o passing lanes as well as me ging a ea leng hs, which
ep esen a c i ical elemen in geome ic design o passing lanes (Ca iso e al., 2018). In addi ion, esea ch
conduc ed by Appiah e al. (2020) in es iga ed he isk o le - u n c ash occu ence a a signalized
in e sec ion using a calib a ed a ic mic osimula ion model.
The e a e also app oaches ocusing on pa icula elemen s o he oad en i onmen , such as he s udy by
As a i a e al (2020b), ocusing on po en ial con lic s wi h oadside objec s. The au ho s de eloped a
dedica ed add-on o enable he es ima ion o new oad sa e y indica o s ha can be used wi h T i one,
Visum and Aimsun, u ilizing he ‘Zombie d i e ’ me hodology o assess oad sa e y le els. The co e o he
me hodology is epo ed o be based on ea lie ideas ha include economical assessmen o isks posed
by he a ious oadside objec s o obs acles, highligh ing dange ous loca ions as ones wi h high c ash
ene gy alues, such as ees o sha p u ns ha a e po en ially haza dous o un-o c ash occu ences.
The ex ensi e knowledge al eady gained on he in luence o he oad geome ic con igu a ion on oad
sa e y isk allows he inco po a ion o sa e y componen s in o design so wa e. The implemen a ion o oad
assessmen in BIM oad models, o ins ance, will no only end in be e -designed oads bu also accoun
o he comple e li e-cycle o in as uc u e (Cos in e al., 2018; S a ý e al., 2019).
Howe e , e en hough SSAM is a widely used ool, he e a e some limi a ions ha should be conside ed
be o e ca ying ou he sa e y analysis. One o he majo p oblems is ha ehicle ajec o y, ha is used
in SSAM o iden i y con lic s and gene a ed by mic osimula ion models, do no e lec complica ed d i e s’
beha iou s in he eal wo ld (Caliendo & Guida, 2012; Huang, Liu, Yu, & Wang, 2013). This is because
cu en CF a e no ealis ic enough o p oduce ealis ic single ajec o ies as s a ed in he p e ious sec ion.
Ano he limi a ion o SSAM is ha only con lic s be ween ehicles a e iden i ied. Vulne able oad use s
such as pedes ians, cyclis s and e-scoo e s o he sa e y impac o buses a e no conside ed (Caliendo
& Guida, 2012). And his is an a ea o esea ch ha gaining mo e impo ance, especially a e he
inco po a ion o in oduc ion o connec ed and au onomous ehicles (CAVs) in mic osimula ion models.
Also, ano he limi a ion o SSAM is he h eshold alues ha a e de ined o he TTC and PET o iden i y
con lic si ua ions. These h eshold alue should be di e en o di e en in e ac ions. Fo ins ance, he e
should no be he same h eshold alue o he in e ac ion be ween human-d i en ehicles o be ween
CAVs (Weije ma s e al., 2021). In ac , se ing hese alues could also induce ce ain p oblems due o i s
main subjec i i y.
Some au ho s (Wang & S ama iadis, 2013; As a i a e al., 2019) a gue ha he e a e s ill no app op ia e
me hodologies and eliable su oga e indica o s o simula ion-based sa e y s udies. In ac , as
mic osimula ion models a e de eloped o a oid c ashes wi h o he ehicles, models p oduced
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
55
inaccu acies ha led SSAM o iden i y con lic s when he TTC was o be ze o (Caliendo & Guida, 2012).
In his case, e en s in SSAM wi h TTC ze o had been excluded om he a ic con lic analysis. An
app op ia e me hod and new simula ion-based su oga e indica o s need o be de eloped and alida ed
(Wang & S ama iadis, 2013).
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
56
Re e ence
Aim o he s udy
Me hodology
Pe o mance/Sa e y
indica o
Simula ion so wa e
D awbacks /
Fu he esea ch
Ge man e
al. (2008)
De elop and e alua e a
me hod o he su oga e
sa e y assessmen o
a ic acili ies.
Vehicle ajec o y da a om
a ic models a e used as
inpu da a o SSAM. Then
su oga e measu es a e
calcula ed co esponding o
each eh- o- eh in e ac ion.
TTC
PET
SSAM can combine
wi h Aimsun,
Pa amics, Texas,
Vissim.
Need o esea ch he
de elopmen o a
composi e sa e y index
ha could ac o in he
mul i ude o
con adic o y
sa e y indica o s.
Need o inco po a e mo e
con lic ypes (no jus he
h ee ypes conside ed:
ea end, lane changing o
c ossing).
Caliendo &
Guida
(2012)
P o e he po en iali y o
mic osimula ion models
o assessing sa e y a
u ban unsignalized
in e sec ions
Unsignalized u ban
in e sec ions om 2003-
2007 we e s udied. T a ic
lows we e measu ed by
ideo came as.
Mic osimula ion model was
used o model a ic low.
Calib a ion & alida ion
p ocesses we e also done.
SSAM was used o iden i y
c i ical con lic s
TTC
PET
Combina ion o
Aimsun and SSAM
Illegal o ms o d i e s’
beha iou
a e no
conside ed in he
analysis.
Se e i y o c ashes has
no been e alua ed.
Mo e esea ch on
con lic s wi h buses,
cyclis s, and pedes ians.
Hab emich
ael e al.
(2014)
E alua e sa e y
implica ion o agg essi e
d i ing by using a
mic oscopic a ic
simula ion app oach.
T a ic simula o p o ides
ajec o y in o ma ion o all
ehicles o SSAM, which
de ec s ehicle con lic s.
Agg essi e d i e s a e
modelled in VISSIM by
changing d i e beha iou
pa ame e s.
C ash isk (odds
a io)
Se e i y le els o
c ashes
PET
Magni ude o
pe cei ed bene i s o
agg essi e d i ing
Combina ion o
VISSIM and SSAM
The me hod does no
include single ehicle
c ashes agai
ns s a ic
elemen s o oad
in as uc u e.
Fu u e wo k should ocus
on use o na u alis ic
d i ing s udies.
Table 3-2 Summa y o e iewed ajec o y-based sa e y assessmen publica ions
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
63
S a
Ra ing
Vehicle
occupan
S a Ra ing
Sco e
Mo o cyclis
S a Ra ing
Sco e
Desc ip ion o Typical A ibu es
Di ided Road Undi ided Road
2-S a s
12.5 o <
22.5 12.5 o < 22.5
Majo de iciencies gi en he
p e a
iling speed include poo
median p o ec ion, poo
oadside condi ions and/o
poo ly designed in e sec ions
a equen in e als.
Majo de iciencies include
poo oadside condi ions,
na ow lanes, and/o poo ly
designed in e sec ions a
equen in e als.
1-S a 22.5+ 22.5+
Gi en he p e ailing speed,
c i ical de iciencies include
poo alignmen , na ow lanes,
na ow shoulde s, poo median
p o ec ion, se e e oadside
haza ds, and nume ous
in e sec ions.
The p e ailing speed, poo
alignmen , na ow lanes,
n
a ow shoulde s, no
median p o ec ion, se e e
oadside haza ds, and
nume ous in e sec ions a e
c i ical de iciencies.
Table 3-5 Typical cha ac e is ics o 1 o 5 s a s sec ion o ehicle occupan s and mo o cyclis s
S a Ra ings a e used o calcula e he collec i e isk o he oad co ido o ne wo k. Fa al and Se ious
Inju y (FSI) Es ima es a e one o he iRAP p o ocols o auma es ima ion. I d aws on he oad a ibu es
used wi hin he S a Ra ing Model oge he wi h low da a o each oad use and ne wo k-le el c ash da a
o es ima e how many a al and se ious c ashes a e likely o occu along each oad segmen o e a 20
yea pe iod.
Applied ac oss he ne wo k, his app oach enables an es ima e o he numbe o ehicle occupan s,
mo o cyclis s, bicyclis s, and pedes ians’ a ali ies on e e y 100-me e sec ion o oad. Wi h he alloca ion
o a ali ies comple ed o a 100-me e s le el, one can also assess he bene i s o ea men s a a 100-
me e sec ion le el. Figu e 3-6 p esen s an example o he FSI es ima ions dis ibu ed on he pa o he
Slo akia ne wo k. This collec i e isk modelling is hen used o suppo he ecommenda ion, p io i iza ion
and cos -bene i analysis o oad sa e y upg ades (iRAP, 2014c).
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
64
Figu e 3-6 Example o FSI es ima es dis ibu ion o Slo akia (Figu e sou ce: iRAP ViDA sys em)
iRAP S a Ra ings ha e been used ex ensi ely a ound he wo ld, as well as in he ollowing Eu opean
p ojec s:
• EU - Sa e yNe P ojec : Based on he knowledge ha oad in as uc u e design and layou
a ec he le el o sa e y o a pa icula ne wo k, he Sa e yNe p ojec de eloped Sa e y
Pe o mance Indica o s (SPI) o allow benchma king among Eu opean coun ies. The p ojec
pilo ed de elopmen s in he Ne he lands, Po ugal, G eece, and Is ael. Ul ima ely, hey
p esen wo ypes o SPIs: a oad ne wo k SPI and a oad design SPI (Yannis e al., 2013).
The oad design SPI elied on da a om he Eu opean Road Assessmen P og amme
(Eu oRAP) and was analysed using he Eu oRAP Road P o ec ion Sco e, a p elimina y
e sion o he S a Ra ing Sco es model which only p o ided sa e y e alua ion o ehicle
occupan s.
• EU- SLAIN P ojec : The SLAIN P ojec aimed o assess he TEN-T co e ne wo k and
es ima e he sa e y pe o mance o oads in 4 coun ies: C oa ia, I aly, G eece and Spain.
Mo e han 4,000km o oad we e assessed as pa o he p ojec (Hol oyd, 2021).
• In e eg - SABRINA P ojec : The p ojec SABRINA aimed a p o iding sa e bicycle ou es
in he Danube a ea. SABRINA's Wo k Package T1 aims o S a Ra e bicycle ou es o sa e y.
The ocus is on p e-iden i ied Eu oVelo bicycle ou es (6,8,9,11,13,14) in he Danube A ea.
App ox. 3500 km o bicycle ou e we e inspec ed using specially equipped ehicles, so wa e
and ained analys s. Inspec ions ocused on mo e han 30 oad design ea u es known o
in luence c ash likelihood and se e i y o inju ies o cyclis s in he mixed a ic en i onmen
as hey pose a highe sa e y h ea o he cyclis s. Following he comple ion o he ideo-based
inspec ion, each ele an design ea u e was measu ed and a ed acco ding o he exis ing
Road Assessmen P og amme (RAP) p o ocols. When he oad inspec ion and a ing p ocess
was comple ed, S a Ra ings we e p oduced o elucida e he isk alloca ion along he ou es
(U sachi e al., 2022). In he SABRINA P ojec , he bicyclis s' ou es we e also analysed using
he Eu opean Ce i ica ion S anda d (ECS) de eloped by Eu opean Cyclis s Fede a ion (ECF)
and he CycleRAP me hodology below:
iRAP has de eloped CycleRAP, an e idence-based in as uc u e isk e alua ion model speci ically o
bicyclis s and ligh mobili y. The model ocuses speci ically on he ea u es o acili ies and he inhe en
isk hey pose ac oss a ange o bicyclis and ligh mobili y c ash ypes, i espec i e o he acili y ype (o
whe he i is pa o he oad ne wo k). Unlike S a Ra ings, CycleRAP can be used on cycling in as uc u e
independen o oad ne wo ks. Fu he mo e, he model accoun s o a b oade ange o c ash ypes,
including bicycle-bicycle, bicycle-pedes ian o some kinds o single bicycle c ash isk (iRAP, 2022).
3.1.11.2 The Eu opean Ne wo k Wide Assessmen (NWA)
The Eu opean Commission ecen ly p esen ed i s me hodology o ne wo k-wide assessmen s on a
me hodological and implemen a ion handbook (Eu opean Commission, 2023). NWA was c ea ed o a
speci ic pu pose which is e alua ion o he sa e y o he Eu opean TEN-T and p ima y oad ne wo k in line
wi h he Road In as uc u e Sa e y Managemen (RISM) Di ec i e equi emen s. I acili a es bo h eac i e
and p oac i e oad in as uc u e sa e y assessmen . I is ecommended o applica ion ou side u ban
a eas, excep o u ban mo o ways. I conside s he p esence o ulne able oad use s, such as bicyclis s
and pedes ians. D awing on he S a Ra ings me hodology, NWA was de eloped o e alua e he
Eu opean TEN-T p ima y oads ne wo k in line wi h he equi emen s o he RISM di ec i e.
The applica ion is di ided in o se e al s eps, s a ing wi h selec ing he oad ype. Da a is collec ed in wo
s ages. The i s s age e e s o o e iew da a, which aims o suppo oad segmen a ion in homogeneous
sec ions o ixed-leng h sec ions. The second s age, a e comple e segmen a ion, seeks o inc ease he
le el o de ail, wi h da a collec ed by sec ion. Fo he nine elemen s wi h da a collec ed on phase 2, c ash
modi ica ion ac o s (CMF) and educ ion ac o s a e es ima ed and used o calcula e he p oac i e sco e.
The model p esen s a mul iplica i e s uc u e as he iRAP model bu ames he le el o isk in h ee bands:
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
65
low, in e media e, and high isk, and does no acili a e any a ali y es ima ion modelling p ocedu es. The
p oac i e NWA me hod does no ake in o accoun he e ec s o speeding ela ed o he in as uc u e in
buil sa e y since i does no conside he ope a ing speed a iable. .The o e all sec ion a ing is calcula ed
based on bo h eac i e and p oac i e assessmen and a ed in i e le els o isk (Ve y High p io i y o Ve y
Low P io i y) o he pu pose o epo ing acco ding o he 2019 RISM Di ec i e NWRSA equi emen s.
(Eu opean Commission, 2023)..
3.1.11.3 The Aus alian Na ional Risk Assessmen Model (ANRAM)
ANRAM is a FSI es ima ion p o ocol calib a ed speci ically o Aus alia. I uses he iRAP S a Ra ing
model inpu da a o p o ide es ima es o a al and se e e inju ies o a ypical i e-yea pe iod based on
six oad s e eo ypes, sec ion leng h and AADT (Aus oads, 2014). The model elies on Sa e y
Pe o mance Func ions (SPF) es ima ed using c ash- eco ded a iables o he Vic o ia egion om 2005
o 2009, combined wi h oad in en o y and AADTs. I only accoun s o ehicle occupan c ashes ( ha is,
no o he oad use g oups) and is speci ic o he Aus alian oad classi ica ion sys em and geome ic
cha ac e is ics (Aus oads, 2020).
3.1.11.4 The In as uc u e Risk Assessmen Model o New Zealand (IRR)
As wi h he NWA and ANRAM, he In as uc u e Risk Ra ing (IRR) de eloped by Waka Ko ahi NZ
T anspo Agency used he iRAP S a Ra ings as s a ing poin s o de ine oad ea u es and ope a ional
a ibu es well as he coding guideline. Ye , he IRR model is a simpli ied model conside ing only eigh
a ibu es. The a ibu es we e de ined by hei con ibu ion o isk and he po en ial o code au oma ion
based on da a sou ces like ae ial image y, oad asse managemen da ase s and s ee iew image y.
The esul s a e he o e all classi ica ion o he indi idual le el o isk (Waka Ko ahi NZ T anspo Agency,
2022). In addi ion o isk scanning, he ool has also been used as a speed managemen ool (Amoh-
Gyimah e al., 2019). No applica ion in u ban a eas o o VRU’s sa e y as iden i ied.
Table 3-6 p esen s a summa y o he ele an a ibu es o he oad sa e y assessmen s me hodologies
e iewed.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
66
Me hodology
iRAP
S a Ra ings,
FSIs and SRIP
NWA
ANRAM
IRR
Type o
app oach
Reac i e/
P oac i e
Reac i e/
P oac i e
P oac i e
P oac i e
Numbe
a ibu es
conside ed
56
22
42
8
VRU’s
iden i ica ion o
isk
Di ec ly
Indi ec ly
No
No
Applica ion in
u ban a eas
Yes
Only in u ban
mo o ways.
Yes. U ban
oad ypes
conside ed in he
analysis
Yes. U ban
en i onmen
conside ed in he
land use a ibu e.
Ou comes and
le els o isk
iden i ied
5 le els o isk (1
o 5 s a s)
Collec i e and/o
indi idual isk
5 le els o isk
(Ve y High p io i y
o Ve y Low
P io i y)
Indi idual isk
FSI c ashes ( o a
i e-yea pe iod)
o a gi en oad
sec ion
Collec i e and/o
indi idual isk
5 le els o isk
(Low isk o High
isk)
Indi idual isk
Table 3-6 Summa y o oad sa e y assessmen me hodologies e iewed.
3.1.11.5 The academic pe spec i e o in as uc u e assessmen s
In an ea lie s udy, G ego iades (2007) p oposed a sa e y p edic ion ea ly wa ning sys em o allow mo e
ime o au ho i ies o eac o highly unsa e/ isky si ua ions. The aim o he s udy was o enhance
adi ional agen -based simula ion wi h Bayesian Belie Ne wo ks (BBN) and, as a esul , imp o e c ash
p obabili y p edic ions. To achie e his, he au ho exploi ed eal- ime obse a ions om a simula ed
ne wo k, enhanced hem wi h a mul i-agen model ha p o ided eal- ime in o ma ion on he s a e o he
ne wo k and in oduced c ash scena ios ia Mon e Ca lo simula ion. The BBN model was ul ima ely
de eloped by exploi ing da a including a ic olumes, oad ne wo k cha ac e is ics, wea he condi ions,
and d i ing beha iou s, among o he s, and he applica ion was cha ac e ized as success ul o e all.
Fu he mo e, G ego iades e al. (2010) expanded on his esea ch di ec ion wi h he in en o p o iding
alida ion o sa e y equi emen s o conside ed oad sa e y designs o ne wo ks. This is achie ed once
again by in eg a ing BBN echnologies on agen -based simula ions. A no ewo hy s eng h o BBN is ha
i can be in o med by p io knowledge, whe he empi ical o om es ima ions based on pas li e a u e. The
condi ional p obabili y ables exploi ed by he model we e popula ed by a iables ha we e signi ican
p edic o s in co ela ion wi h pas c ash eco ds. The de eloped Road Sa e y Analyse (RoSA) ool
achie ed sa is ac o y pe o mance, hough he au ho s no ed ha u he e inemen s we e needed.
Rega ding pa icula in as uc u e elemen s, mic osimula ion models can become inc easingly composi e
o e lec he elemen o en i onmen unde conside a ion. Speci ically, Tan e al. (2012) de eloped a
mic oscopic simula ion model o assess he sa e y o signalized in e sec ions. To c ea e i , a numbe o
key beha iou al aspec s o oad use s we e conside ed, such as he s op–go decision a he onse o a
yellow ligh , u ning pa hs, u ning speeds, s a -up esponse ime, and pedes ian gap accep ance. These
aspec s we e independen ly de eloped as simula ion sub-models and in eg a ed in he o e all model,
while limi ed alida ion wi h a eal in e sec ion enhanced by image p ocessing also ook place in he s udy.
O he in as uc u e elemen s ha e been also conside ed om di e en scopes, such as by he s udy o
Tes a e al. (2022) who in es iga ed ways o ob ain alues o s uc u al loads om a ic o he
assessmen o b idges. The au ho s u ilized mic oscopic simula ion o ob ain OD alues in o de o explo e
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
67
he s uc u al sa e y o b idges, using pay- oll da a and egional egis a ion da a as simula ion inpu s,
along wi h ce ain assump ions o a ic beha iou . While his is no a s ic ly c ash-based o oad sa e y
assessmen pe se, he a ic simula ions employed in his wo k p o ided he means o sol ing his
ne wo k-le el issue.
3.1.11.6 Coun ies o oad ype-speci ic me hodologies
The li e a u e b ings a se ies o ini ia i es o de eloping oad sa e y assessmen s me hodologies ha a e
coun y speci ic. In Eu ope, hey ha e he o m o C ash P edic ion Models and can be ound o I aly (la
To e e al., 2019), Czech epublic (Amb os e al., 2019), Po ugal (da Cos a e al., 2018) and o he . Mos
o hese models a e also de eloped o se o oad ypes like eeways o mo o ways (Amb os e al., 2019;
la To e e al., 2019) and na ional oads o highways (Amb os e al., 2019; da Cos a e al., 2018).
Using he a ic con lic echnique when using mic oscopic simula ion, se e al sa e y implica ions we e
e alua ed o highways in speci ic. Fo ins ance, a s udy conduc ed by Hab emichael and De Picado
San os (2014) in es iga ed agg essi e d i ing in mo o ways o di e en a ic condi ions using simula ed
con lic s o de e mine c ash- isk, and Pos Enc oachmen Time (PET) and a el ime o de e mine he
se e i y le els o he expec ed c ashes. A simila s udy explo ed eeway in e change me ging a eas by
ob aining he hou ly composi e isk indexes as well as de eloping a mul i a ia e linea eg ession model
(Li e al., 2016). In addi ion, a ecen s udy in es iga ed speci ically he in luence o he le ha d shoulde
on he sa e y o ehicles a eling on mul i-lane highways, based on con lic cha ac e is ics unde di e en
numbe s o lanes (Zhao e al. 2022a). A c i ical cha ac e is ic o eeways is managed lanes, which can
imp o e a ic mobili y and gene a e e enue o anspo a ion agencies. The e o e, esea ch also
ocused on speci ying he sa es accessibili y le el and iden i ying he sa es wea ing leng h nea access
zones based on simula ed a ic con lic s along wi h a con lic equency analysis (Saad e al. 2018).
Ano he ele an simula ion s udy explo ed he a ic beha iou and ehicle in e ac ions in a oad
in e change o highway by quan i ying he impac s o speed limi a ia ions on bo h a ic and sa e y
(Ribei o e al. 2019). Finally, he impac s o low-speed ehicles on exp essway a ic sa e y has been also
examined and he cha ac e is ics o he a ec ed ehicles we e de ined in a s udy conduc ed by Xu e al.
(2022).
Focusing on u al oad en i onmen s, he po en ial e ec s o an o e aking assis an o wo-lane u al
oads we e iden i ied h ough a mic oscopic a ic simula ion s udy by Hegeman e al. (2009), indica ing
ha a sa e y imp o emen can be accomplished wi hou nega i e consequences o a ic e iciency and
d i e com o .
3.1.11.7 Risk assessmen s o u ban en i onmen s
The a ic mic oscopic simula ion app oach was used on in es iga ing he isk assessmen o u ban
ne wo ks as well. Due o a ic conges ion no iced in u ban en i onmen s, signal se ing and mul iple
u ning-lane assignmen a in e sec ions was explo ed in a mic oscopic simula ion s udy conduc ed by Li
and Sun (2019) aking in o accoun oad sa e y based on ehicle con lic s and pedes ian in e e ence.
Ano he sa e y issue o u ban en i onmen s ela ed o d i ing beha iou is ha d i e s un ed a ic ligh s
a signalized in e sec ions. The e o e, esea ch e o s ha e been made in o de o in es iga e he possible
causes and e ec i eness o coun e measu es especially when c ash da a a e no a ailable o eliable o
s a is ical analysis. Lee e al. (2018) p oposed a a ic simula ion con lic -based app oach o e alua ing
he ed ligh unning coun e measu es and es ed he mos p e alen scena ios, which we e: inc easing
he yellow signal in e al du a ion, ins alling an ad ance wa ning sign, and a came a. Mo eo e , a di e en
simula ion s udy explo ed he sa e y impac s o a p o ec ed in e sec ion design o bicyclis s concluding
ha imp o es oad sa e y (P es on and Pulugu ha, 2022; Mon ella e al., 2020; Ricca di e al., 2022).
Wi h ega ds o public anspo , a mic oscopic a ic simula ion modeling app oach was also adop ed in
a s udy conduc ed by Goh e al. (2014) in o de o quan i y he oad sa e y e ec s o implemen a ion o
bus lanes. On a simila no e, Kapa ias e al. (2020) wished o ex end a p e iously c ea ed p edic i e
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
68
e alua ion ool, co e ing pollu ion and a ic e iciency, named CONDUITS_DST. The ex ension aimed o
ake in o conside a ion co ela ions o a ic cha ac e is ics o oad sa e y indica o s. As a case s udy, a
bus p io i y signaling sys em in he ci y o B ussels was in es iga ed. The simula ion models had been
es ed based on da a om simula ion models ob ained be o e and a e implemen ing he signaling
sys em. Bo h he segmen le el and he ne wo k le el we e conside ed. The au ho s no e a sa is ac o y
pe o mance o he ool ex ension, which de ec ed bo h posi i e and nega i e e ec s o he public
anspo p io i y signaling sys em in eg a ion in he u ban cen e o B ussels.
Olmez e al. (2021) no ed he mo e equen speeding ins ances in u ban en i onmen s. Thei s udy was
hus incen i ized o exploi agen -based mic osimula ion modelling and o co ela e collisions wi h
speeding and a ic densi y pa ame e s. The au ho s used au onomous agen s o achie e an o e all
he e ogeneous global sample. In addi ion, he au ho s conside ed he u ili y o highe a ic densi y as a
speeding de e en ool o educe c ashes caused by speeding and hus inc ease oad sa e y le els. The
s udy sligh ly de ia es om he li e a u e in a sense ha collisions inc ease disp opo iona ely highly as
a ic densi y inc eases a low and modes a ic densi ies; howe e , collisions beha e mo e
p opo iona ely a e a c i ical poin in mid- ange densi y. Mo eo e , ehicles adhe ing close o speed
limi s we e ound o linea ly educe collisions.
Ano he s udy (Pe o e al., 2020) p oposed a p i acy ensu ing eme gency ehicle app oaching wa ning
sys em (PEEV-WS), which was de e mined as a sub-ca ego y o Vehicula Ad hoc NETwo ks (VANETs).
An in e es ing sub ask in ol ing mic oscopic simula ion was c ea ed in o de o de e mine whe he PEEV-
WS p o ided simula ed d i e s wi h su icien eac ion imes, which was con i med o be he case. An
ea lie s udy had ollowed a simila ein, in es iga ing he a-p io i in oduc ion o Ad anced D i e
Assis ance Sys ems (ADAS) h ough mic osimula ion (Lundg en & Tapani, 2005). The au ho s de eloped
a longi udinal con ol pa o he d i ing ask desc ibed by a ca - ollowing mic oscopic simula ion model;
hey concluded ha ADAS-induced beha iou al changes a e impo an o conside while moni o ing oad
sa e y le els.
3.1.11.8 Road sa e y assessmen o au oma ed mobili y
Since he e is lack o his o ical gene alizable c ash da a especially in case o high ma ke pene a ion
a es o Connec ed and Au oma ed Vehicles (CAVs), mic oscopic simula ion me hod is conside ed as an
ideal app oach o s udying au oma ed mobili y impac s on sa e y. Fo his eason, esea ch has been also
ocused on his di ec ion. A e iew o se e al ecen mic osimula ion s udies in CAV sa e y impac s based
on su oga e measu es o sa e y, wi h ocus on highway segmen s, can be ound in Papadimi iou e al.
(2022). This e iew concludes ha while mos s udies indica e subs an ial sa e y imp o emen s by highe
a ic pene a ion a es o CAVs, he simula ion me hodologies and speci ic me ics and hei h esholds
a y conside ably among s udies, and a common amewo k is lacking.
In pa icula , a mic osimula ion s udy conduc ed by Elawady e al. (2022) in es iga ed he impac o CAVs
on in e sec ion a ic sa e y using he SSAM unde di e en pene a ion a es o CAV condi ions.
Addi ionally, Jeong and Oh (2017) p oposed a me hodology o assess he e ec i eness o ac i e ehicle
sa e y sys ems (AVSSs) ha co espond o Le el 2 ehicle au oma ion using a mic oscopic a ic
simula o and SSAM o de i e indi ec sa e y measu es. A simila esea ch which ocused on alida ing
simula ed con lic s, applied SSAM as well, o a case s udy o ecen ly de eloped CAV applica ion (Essa
and Sayed, 2020).
An ongoing challenge acing he en i e anspo sec o is he in eg a ion o CAVs oge he wi h
con en ional, human-d i en a ic and Vulne able Road Use s (VRUs) such as pedes ians and cyclis s.
Koopmann e al. (2022) sough o in es iga e his issue by conside ing an in elligen con olle ins alled
on an u ban in e sec ion in a mic oscopic simula ion scena io, which would in eg a e beha iou al ai s o
con en ional o au oma ed mobili y and all o he oad use ypes. The de eloped model ea u ed a
complex a chi ec u e pe oad use ca ego y, while also being enhanced by game heo y applica ions o
be e exp ess he in e ac ions be ween oad use s while nego ia ing passage h ough he in e sec ion.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
69
Resul s a e men ioned o be ye p elimina y and qui e demanding in implemen a ion e o s, wi h possible
ma gins in a ic e iciency.
In case o connec ed ehicles, he e is a con inuous s eam o loca ion da a suscep ible o acking a acks.
To ha end, a mic oscopic a ic simula ion amewo k was de eloped, syn hesizing se e al silence-
based loca ion p i acy schemes by Xin e al. (2019). Ne e heless, he e is a c i ical elemen wi h ega ds
o au oma ed ehicle con ol sys ems elying on he use o ei he onboa d ange senso s o ehicle- o
ehicle wi eless communica ions, which is he delay in da a acquisi ion. Hence a mic osimula ion s udy
by Liu e al. (2006) a emp ed o assess he sa e y impac s o delayed in o ma ion o in elligen ehicle
con ol sys em applica ions and highligh ed ha he ehicle con ol sys ems ope a ion is signi ican ly
a ec ed. Mo eo e , he s eadily inc easing o CAV le els is a ac in ansi se ices as well, and hence
insigh s abou ansi se ices sa e y such as au oma ed poin - o-poin shu le bus se ices (Oikonomou
e al., 2020) and au onomous on-demand mobili y se ices (Mou akos e al., 2021) ha e been also gi en
based on simula ed a ic con lic s.
I is wo h highligh ing ha connec ed mobili y migh enable he calcula ions o new indica o s, o o
acili a e applica ion oppo uni ies o exis ing ones, such as he c ash po en ial index (CPI) compu ed in a
s udy by Jo e al. (2021) wi h da a ha we e also compa ible wi h DRAC calcula ion. Fu he mo e, as he
adi ional su oga e sa e y me ics ha e been mos ly es ablished o con en ional ehicles, he e a e
many di icul ies when hey a e used o he assessmen o au oma ed ehicles. Fo his eason, p oac i e
me ics we e used ins ead by Ma as e al. (2019). Speci ically, he p oac i e uzzy su oga e sa e y me ic
(PFS) and he c i ical uzzy su oga e sa e y me ic (CFS) o ea end collisions ha e been de eloped
and he esul s indica ed hei obus ness on e alua ing he sa e y le el in he longi udinal di ec ion. These
uzzy su oga e sa e y me ics we e u he used in a di e en s udy by Ma as e al. (2021), in which a
uzzy con olle was de eloped o adap i e c uise con ol (ACC) and alida ed using eal-wo ld da a and
a ic simula ion. In addi ion, he exis ing models used o simula e ehicula a ic a e conside ed as no
alid o p edic a ic lows unde inc easing amoun s o ehicle au oma ion. Hence in a s udy by an Lin
e al. (2016), an ad anced open-sou ce simula ion amewo k named OpenT a icSim was p oposed,
o e ing he inc emen al ex ension o mic oscopic models wi h explana o y men al models, and going one
s ep u he o he nex gene a ion o a ic simula ion models ha a e needed in he coming decades.
3.2 Modal shi models
The li e a u e was e iewed in wo ca ego ies: mode choice modelling wi h he Random U ili y Models
(RUM), and mode choice modelling wi h Machine Lea ning models.
3.2.1 De ini ion o modal shi
The concep o modal shi ac s as a basis o he ollowing sec ions. The exp ession "Modal Shi (MS)"
is de i ed om he e m "Mode Choice (MC)". MC is he decision-making p ocess by indi iduals o pe son
g oups (o can also be e med as a el beha iou ) while choosing a mode o anspo (e.g., Bike) om a
se o a ailable al e na i e modes (e.g., Bike, Walk, Ca o Public anspo (PuT)) o a el om he o igin
(O) o he des ina ion (D) (D. A. Henshe & Bu on, 2007; D. A. ; Henshe & Bu on, 2008; O úza &
Willumsen, 2014) o a ious ip pu poses (e.g., Shopping, Commu ing, Educa ion o leisu e). Fac o s
like a el ime, cos , con enience, com o , sa e y and en i onmen al conce ns signi ican ly in luence
mode choice decisions (Ilahi e al., 2019; Li e al., 2021; McCa hy e al., 2017; Rahul & Ve ma, 2013).
Ma hema ically, mode choice is exp essed in e ms o he p obabili y (o pe cen age) o using one
pa icula mode (e.g., Bike) ou o all he a ailable modes o anspo (e.g., Bike, Ca and PuT).
The MS e e s o he change in he p obabili y o pe cen age (ei he an inc ease in he p obabili y o a
dec ease om he base alue) o using a speci ic mode ou o all he a ailable modes o anspo by
indi iduals o pe son g oups o di e en pu poses (Ba y e al., 2015). MS may a ise wi h o wi hou
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
70
a i icial (human) in e en ion, e.g., a i icially, MS may be induced (by changing ac o s like a el cos ,
ime, con enience e c.) o achie e pa icula objec i es like:
• Reducing conges ion
• Reducing ca bon oo p in
• Imp o ing ai quali y
• P omo ing sus ainabili y
Fo example, an MS owa ds ac i e modes om he mode ca can educe he ca bon oo p in subs an ially
(Ba y e al., 2015; Lude e e al., 2021). Na u ally, MS may occu due o changes in economic and social
condi ions (e.g., highe economic p ospe i y may induce an MS owa ds he mode ca om he mode PuT)
(Puche , 1994). A e a b ie de ini ion o MC and MS, he me hods and me hodologies o modelling MC
and, subsequen ly, he me hods and me hodologies o inducing MS will be discussed in he ollowing
sec ions.
3.2.2 Modelling Mode Choice (MC)
MC deals wi h he choice o a pa icula mode by a use o use g oup ou o all o he a ailable modes.
Classically, hese ypes o p oblems a e modelled wi h choice modelling echnique. Al hough ini ially
agg ega ed models we e used o model o p edic he MC. These models o app oaches ocused on he
choices o he modes made by an a e age indi idual o he ips be ween an o igin and des ina ion
(Domencich & McFadden, 1975). These app oaches we e:
• non-beha iou al and eplica ed he esul s acco ding o he condi ions exis ing du ing he
su ey.
• no policy-o ien ed.
• based on he agg ega ion o he su ey da a.
To o e come hese p oblems, disagg ega ed modelling app oaches gained popula i y du ing he 1960s.
The da a o his modelling ypology we e collec ed a he indi idual o household le el. Simila ly, he
pa ame e s o hese models we e also es ima ed ac oss he indi idual o household le el (Ba e al.,
1982). S a is ically, he dependen a iables o he mode choice models a e disc e e and ini e (e.g., i he
mode is aken(1) o no (0)). In con as , he independen a iables a e gene ally con inuous (M. E. Ben-
Aki a & Le man, 2006). Ne e heless, he cu en echniques o MC modelling can be classi ied unde
wo majo ca ego ies, namely, Random U ili y Model (RUM) and Machine Lea ning Models (MLM). This
classi ica ion is explained in Figu e 3-7.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
71
Figu e 3-7 Me hodologies o mode choice modelling
1. Model choice and se -up: An in-dep h iden i ica ion and ca ego isa ion o he anspo modes
a e ca ied ou o de e mine he model class and ype o be used o MC modelling. A sui able
analy ical (choice) model (some examples o analy ical models used o MC modelling a e Logi ,
P obi o Nes ed Logi ) is hen chosen based on he iden i ica ion and ca ego isa ion o he
anspo modes. This model is hen app op ia ely se up (s eps like he decla a ion o he a iables
and a iable ypes) o ini ia e he modelling p ocess.
2. Model es ima ion and assessmen : Once he i s s ep is o e , he model is es ima ed wi h he
help o he s a ed p e e ence su ey da a. A e he es ima ion p ocess, he MC model can
p oduce esul s (in e ms o p obabili ies o pe cen ages) o e e y mode usage. These
p obabili ies a e also known as he mode sha e o he espec i e modes.
3. Model alida ion and calib a ion: The es ima ed MC model's ini ial esul s o ou pu s o en di e
as ly om he eal-wo ld scena io (obse a ions). Thus, i is alida ed by compa ing he ou pu s
wi h eal-wo ld obse a ions (da a). The model is calib a ed by al e ing he es ima ed pa ame e s
and un again o gene a e new ou pu s. The ou pu s a e e- alida ed wi h eal-wo ld da a. This
p ocess is esumed ill con e gence is achie ed. This s ep ensu es plausible esul deli e y om
he MC model.
4. P edic ion and sensi i i y analyses: Once he MC model is calib a ed and he model ou pu s
a e alida ed, he model can be used o p edic ing scena io-speci ic mode sha es. The alues o
he inpu a iables may also be changed acco dingly o conduc sensi i i y analyses o he mode
sha es. This s ep is ca ied ou o obse e he e ec i eness o he modi ica ions (e.g.,
modi ica ions in ime o cos ) in a el demand.
The MC modelling p ocess is desc ibed in Figu e 3-8 below.
Deli e able 1.1 – S a e o he A and end-use needs e iew
Copy igh © by PHOEBE
phoebe-p ojec .eu
72
Figu e 3-8 P ocess o MC modelling. Re ised om (Pes el e al., 2016)
3.2.2.1 Random U ili y Models o Me hods (RUM)
RUMs a e he adi ional models o me hods o modelling he choices o indi iduals o g oups om an
a ailable se o disc e e al e na i es (Ho owi z e al., 1994). The p e e ences o he indi idual o a g oup
o a pa icula al e na i e om a se o al e na i es a e desc ibed by a u ili y unc ion o simply u ili y. The
indi idual o a g oup chooses he al e na i e ( om he a ailable se o al e na i es) o which he pe cei ed
u ili y is highes o ha indi idual o ha g oup (Kamaku a e al., 1982). Table 3.8 (Rep oduced om
(Ra ou e al., 2014)) p esen s he main cha ac e is ics o he RUM ( adi ional mode choice modelling
echniques).
3.2.2.2 Machine Lea ning Models o Me hods
T adi ionally, RUMs a e used o model he mode choice o p edic ing a el demand. Due o he clea
ma hema ical and s a is ical amewo k, he Mul inomial Logi Model (MNL) has been widely used o MC
modelling. Howe e , he MNL model ollows he independence o i ele an al e na i es (IIA) assump ion,
i.e., he p obabili y o choosing a pa icula al e na i e om a gi en se o al e na i es is independen o
all o he al e na i es. Ne e heless, mul iple s udies ha e shown ha use s/ a elle s do no ollow his
assump ion du ing choosing al e na i es (modes o anspo ), leading o biased o ecas s. Se e al
disc e e models like P obi , Nes ed Logi o mixed logi models we e p oposed o e ime o o e come
hese p oblems. Howe e , es ima ing hese models is complica ed compa ed o he MNL model (Tamim
Kashi i e al., 2022). Apa om hese p oblems, he e a e o he p oblems wi h he s anda d RUM me hods
o models, e.g., MNL equi es a pa icula ype o inpu da a s uc u e which, i dis u bed, may p oduce
e oneous ou pu s.
On he con a y, ML modelling me hods a e no based on s ic p io assump ions and allow he compu e s
o p obe he da a s uc u e, ha e a mo e lexible modelling a chi ec u e and o en esul in be e p edic i e
capabili ies (Jamal e al., 2021). In o de o anquish he p oblems o he RUM me hods, ML me hods a e
phoebe-p ojec .eu
Copy igh © by PHOEBE
79
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 3-10 Reasons o he gene a ion o he induced demand
Deco la-Souza & Cohen (1999) is one o he ea lies esea ch wo ks demons a ing induced a el
demand's ole in a highway expansion p ojec (DeCo la-Souza & Cohen, 1999). These indings a e also
suppo ed by o he esea ch s udies (Du an on & Tu ne , 2011; Hymel e al., 2010; Small, 2013). Deco la-
Souza & Cohen (1999) also p o ides he p obable sou ces o induced demand due o in as uc u e
imp o emen s as lis ed below:
• Inc ease in pe son ip p oduc ion (P) ela ed de elopmen .
• Inc ease in pe son ip a ac ion (A) ela ed de elopmen .
• Inc ease in he numbe o daily mo o ised pe son ip P's and A's pe de elopmen uni .
• Inc ease in a e age mo o ised pe son ip dis ance.
• Inc ease in sha e o pe son a el by p i a e mo o ised ehicles.
• Shi in ehicle a el o imp o ed acili ies om unimp o ed acili ies wi hin a co ido o o an
imp o ed co ido due o a ic di e sion om o he co ido s.
The e a e esea ch s udies, ha also ocus on he long- e m aspec s o induced a el demand caused due
o in as uc u e imp o emen s. Ce e o (2003) u ilised da a om 1980 o 1994 o Cali o nia's egion dealing
wi h eeway expansion, a ic olumes, demog aphic, geog aphic aspec s, and ehicle-miles- a elled o
calcula e long- e m elas ici ies (Ce e o, 2003). Noland & Quddus (2006) ound ha inc easing oad a ea
and in oducing a ic signal con ol sys ems o deconges ing he a ic low could induce addi ional a el
demand (Noland & Quddus, 2006). Apa om in as uc u e imp o emen s, speed inc ease also induces
addi ional a el demand (Paglia a & P es on, 2013).
Addi ional a el demand is also gene a ed o induced when a new mode o anspo is appended o he
exis ing sys em. Mul iple esea ch s udies suppo ha ip p oduc ion and a ac ion a e posi i ely a ec ed
by he geog aphic accessibili y a he ends o he ips (Koenig, 1980; Robinson & Vicke man, 1976; Thill &
Kim, 2005). O en he mo i a ion o add a new mode o anspo in o an exis ing sys em is o imp o e
accessibili y, com o and o educe a el ime (Thill & Kim, 2005). Ai mobili y is one such mode ha could
phoebe-p ojec .eu
Copy igh © by PHOEBE
80
Deli e able 1.1 – S a e o he A and end-use needs e iew
induce subs an ial a el demand (Linebe ge e al., 2018). Recen s udies show ha ai mobili y can induce
and p oduce eno mous a el demand whe e he p ices o ai mobili y a e ela i ely low (Balać e al., 2019).
In addi ion o ai mobili y, anspo modes like sha ed mobili y can also induce addi ional a el demand.
Sha ed mobili y p o ides se ices be ween he exis ing public and p i a e modes o anspo a ion. I
p o ides doo - o-doo se ice wi hou ehicle owne ship. (Susan Shaheen e al., 1999). Gi en he inc eased
accessibili y and low p ices, sha ed au onomous ehicles can inc ease a el dis ance and conside ably
induce addi ional a el demand (Ellio & Shaheen, 2016).
3.6.1.2 The need o accoun ing he induced demand
T adi ionally, he a el demand o ecas is ca ied ou based on he ac o s like land use, demog aphics,
income and employmen . These a e exogenous ac o s. While o ecas ing he a el demand h ough a
model based on he ac o s men ioned abo e, he phenomenon o a ic low co ela es s ongly wi h luid
low in he oad ne wo k. Howe e , acco ding o Jacobsen (1997), a ic mo emen s co ela e o he
p ope y o gas o expand in all di ec ions and ill up any a ailable space (Jacobsen, 1997).
The simila i ies o he p ope ies o he a ic mo emen s wi h he p ope ies o gas mani es ha addi ional
capaci y on he supply side (a ailable space) s imula es auxilia y demand. This addi ional a el demand is
known as induced demand. The gene a ion o induced demand can be a ibu ed o he di e en
anspo a ion sys ems' ex eme in e connec i i y and complica ed ela ionships (Casce a, 2009). The
complica ed in e ac ions c ea e a pa adox o a icious cycle be ween he imp o ed in as uc u e and
induced demand (Go ham, 2009). Thus any in as uc u e imp o emen wi hou p ope conside a ion o
induced demand may no gene a e desi able esul s. The Figu e 3-, below ep esen s he icious cycle o
he induced a el demand.
Figu e 3-11 Vicious cycle o induced a el demand (Go ham, 2009)
The induced a el demand may ha e se e al o he epe cussions, including empo a y modal shi s,
empo a y a el speed and ime changes and sa e y conce ns illus a ing he need o model he induced
demand (Lee e al., 1999; Li man, 2017).
phoebe-p ojec .eu
Copy igh © by PHOEBE
81
Deli e able 1.1 – S a e o he A and end-use needs e iew
3.6.1.3 Classi ica ion o he di e en induced demand ypes
In o de o model he phenomenon o induced demand, i is c ucial o ca ego ise he di e en ypes o
induced demand. The ca ego ies unde which he induced demand can be classi ied a e:
a) Ca ego isa ion based on supply side capaci y (Benjamin Schneide , 2018; Li man, 2017)
b) Ca ego isa ion based on he ypes o induced demands (Benjamin Schneide , 2018; Go ham,
2009)
The e a e ou ypes o induced demands acco ding o he ca ego isa ion based on supply side capaci y
(Benjamin Schneide , 2018; Li man, 2017):
a) Gene a ed a ic: Addi ional peak-pe iod ehicle ips on a pa icula oadway occu when capaci y
inc eases. This may include a el ime, ou e, mode, des ina ion and equency shi s.
b) Induced a el: An inc ease in o al ehicle mileage due o oadway imp o emen s ha inc ease
ehicle ip equency and dis ance bu exclude a el shi ed om o he imes and ou es.
c) La en demand: Addi ional ips would be made i a el condi ions imp o ed (less conges ed,
highe design speeds, lowe ehicle cos s o olls).
d) T iple con e gence: Inc eased peak-pe iod ehicle a ic olumes esul when oadway capaci y
inc eases, due o shi s om o he ou es, imes and modes.
Any in e en ion on a selec ed anspo supply elemen may lead o di e en kinds o ypes o induced
demand and impac s (Benjamin Schneide , 2018; Go ham, 2009). These ypes o induced demands can
be ca ego ised as ollows:
a) Di ec induced demand: I e e s o he induced demands ha a e caused by he conscious
decision o he a elle s in he ne wo k. These demands a e c ea ed when a elle s consciously
ake ad an age o he change in a el ime by eo ganising o al e na ing hei a el pa e ns.
These eo ganisa ions can ake place o e di e en empo al ames, e.g., Ins an aneous a elle
esponse (immedia e eo ganisa ion in a el beha iou co esponding o he empo al, spa ial o
modal choice), Sho - un a elle esponse ( eo ganisa ion in a el beha iou co esponding o
hei ip des ina ion, ip chaining pa e ns o equency) and Long- un a elle esponse
( eo ganisa ion in a el beha iou co esponding o hei loca ion choices, e.g., eloca ion o
household o wo k loca ion).
b) Indi ec induced demand: I e e s o he inc emen in a el demand caused by a speci ic
imp o emen in anspo supply.
c) Induced and di e ed demand: I e e s o he inc emen in a el demand ha may ha e occu ed
anyway, i.e., ei he a some o he place o ime in a gi en ne wo k.
3.6.1.4 A ailable modelling and analysis ools and echniques
Analysing and modelling induced a el demand is simila o analysing and modelling no mal a el
beha iou . Disc e e choice modelling is he mos widely used ma hema ical and s a is ical ool o model
induced a el demand. Sec ion 3.2.2 abou he S a e-o - he-A o Modelling Mode Choice, should be
e e ed o o a de ailed analysis and modelling me hodology. In addi ion o DCM, elas ici y analysis is also
used o analyse induced a el demand (DeCo la-Souza & Cohen, 1999).
Howe e , he sub le di e ence be ween induced demand modelling and mode choice modelling is ha he
induced demand is no caused by he common easons (e.g., popula ion, economy e c.) ha cause na u al
a el demand. Thus, i becomes c ucial o iden i y he demand as induced o non-induced demand (Kalliga,
2021).
phoebe-p ojec .eu
Copy igh © by PHOEBE
82
Deli e able 1.1 – S a e o he A and end-use needs e iew
3.4 Socio-economic impac e alua ion
Road a ic c ashes in ol e economic and pe sonal losses and implica e pe sonal inju ies and casual ies.
O en he cos s a ising due o hese losses a e no bo ne by hose who ha e caused hem. The e o e, RTA
can be conside ed an ex e nal cos (El ik, 1994). To e adica e hese ex e nal cos s, on he one hand, a
de ailed knowledge base abou he damages ha RTAs in lic on he economy needs o be acqui ed, and,
on he o he hand, p ope ( oad) sa e y measu es o educe he RTAs need o be in oduced. Such
measu es can be in oduced a di e en s ages o an c ash scena io (Baum e al., 2000). The s ages whe e
hese sa e y measu es may be in oduced a e shown in Figu e 3-12
Figu e 3-12 S ages whe e oad sa e y measu es can be in oduced. Rep oduced om (Baum e al., 2000)
The e exis se e al ins umen s o ensu e oad sa e y. Acco ding o Uni ed Na ions Economic Commission
o Eu ope (UNECE), hese ins umen s co e aspec s like (Lau ina ičius e al., 2012; UNECE, n.d.):
• T a ic ules
• Road signs and signals
• Cons uc ion and echnical inspec ion o ehicles
• Road in as uc u e
• D i ing imes and es pe iods o p o essional d i e s
• Sa e anspo o dange ous goods and haza dous ma e ials
The e a e o he ins umen s as well. The go e nmen needs o in e ene by es ablishing oad sa e y
measu es and policies o ensu e oad sa e y. To es ablish he measu es and policies, he policymake s
(o en he go e nmen ) mus make a clea choice among hese ins umen s o asce ain he highes social
and economic e u ns. The socio-economic analysis is he ool o app aisal me hod o suppo he
es ablished policies and egula o y and in es men p io i ies (Wesemann, 2000). A sound knowledge o he
espec i e (quan i a i e and quali a i e) me hods o he socio-economic analyses is equi ed o conduc
phoebe-p ojec .eu
Copy igh © by PHOEBE
83
Deli e able 1.1 – S a e o he A and end-use needs e iew
such an analysis o oad sa e y measu es. Some o he s a e-o - he-a me hods o conduc ing socio-
economic analysis a e explained in he succeeding sec ions.
Once he selec ed s udies a e ca ego ised, looking in o he a ious me hods deployed o ca y ou socio-
economic analysis becomes necessa y. A de ailed e iew o he selec ed sou ces e eals he p e alen
me hods o ca y ou a socio-economic analysis o oad sa e y measu es and he p incipal sub-me hods o
sus ain he cos and bene i calcula ions o he p e alen me hods. While he "Cos -Bene i Analysis" is a
e y popula commonly used me hod (Baum e al., 2000; OECD, 2001; Wesemann, 2000a). The e a e also
o he me hods o socio-economic assessmen (Baum e al., 2000). The di e en me hods o socio-
economic analysis (and hei sou ces) a e shown inFigu e 3-13, and he ele an s udies a e lis ed in Table
3.11.
Figu e 3-13 Me hods and acili a ing (sub-)me hods o socio-economic analysis o impac s o oad sa e y measu es
phoebe-p ojec .eu
Copy igh © by PHOEBE
84
Deli e able 1.1 – S a e o he A and end-use needs e iew
Me hodology o me hod o
analysis
Sou ces
Cos -Bene i Analysis
(Baum e al., 2000; Daniels e al., 2019; El ik, 2001; Vecino-O iz &
Hyde , 2014; Wesemann, 2000; Wijnen e al., 2009)
Cos -E ec i eness Analysis
(Baum e al., 2000; Chan i h e al., 2021; OECD, 2001; Wesemann,
2000)
Willingness o Pay me hod
(Haddak e al., 2016; OECD, 2001; Rizzi & O úza , 2006)
Mul i-C i e ia P ocesses
(Baum e al., 2000; OECD, 2001; Wesemann, 2000)
Be o e-A e Analysis
(El ik, 2002; Mpogas e al., 2017; Pu e al., 2021; Yanmaz-Tuzel &
Ozbay, 2010)
Table 3-10 Exis ing me hodologies o socio-economic analysis
3.4.1 Cos -Bene i Analysis
Cos -Bene i Analysis (CBA) is he mos used quan i a i e me hod o assess o analyse he socio-economic
impac s o oad sa e y measu es. CBA is a a he sophis ica ed and objec i e way o analysis. CBA aces
i s o igins back o economic wel a e heo y. The cos componen o he CBA is he cos s incu ed in se ing
up he measu e, while he bene i s a e de ined as sa ings h ough measu es, i.e., he educ ion in c ash
cos s. A deployed measu e is mac o-economically p o i able i (i) I he a io o he bene i s (o he measu e)
(BCR) o ha o he cos s (o he measu e) is g ea e han o equal o 1 (Baum e al., 2000; Wesemann,
2000); and (ii)I he ne p esen alue (NPV) is posi i e, i.e., he di e ence be ween he bene i s and he
cos (Daniels e al., 2019).
To calcula e he BCR and NPV, he bene i s' alues a e exp essed mone a ily. Ma hema ically, i can be
exp essed as:
𝐶𝐶𝑠𝑠𝑟𝑟𝑡𝑡−𝐵𝐵𝑡𝑡𝑂𝑂𝑡𝑡𝑓𝑓𝐿𝐿𝑡𝑡 𝑆𝑆𝑟𝑟𝑡𝑡𝐿𝐿𝑠𝑠 (𝐶𝐶𝐵𝐵𝑆𝑆) = 𝐵𝐵𝑡𝑡𝑂𝑂𝑡𝑡𝑓𝑓𝐿𝐿𝑡𝑡𝑟𝑟
𝐶𝐶𝑠𝑠𝑟𝑟𝑡𝑡𝑟𝑟 = 𝑟𝑟𝑡𝑡𝑜𝑜𝑖𝑖𝑠𝑠𝑡𝑡𝐿𝐿𝑠𝑠𝑂𝑂 𝑠𝑠𝑓𝑓 𝑟𝑟𝑠𝑠𝑠𝑠𝐿𝐿𝑜𝑜𝑡𝑡𝑂𝑂𝑡𝑡 𝑠𝑠𝑠𝑠𝑟𝑟𝑡𝑡𝑟𝑟
𝑠𝑠𝑠𝑠𝑟𝑟𝑡𝑡𝑟𝑟 𝑠𝑠𝑓𝑓 𝑚𝑚𝑡𝑡𝑟𝑟𝑟𝑟𝑖𝑖𝑟𝑟𝑡𝑡
𝑁𝑁𝑡𝑡𝑡𝑡 𝑡𝑡𝑟𝑟𝑡𝑡𝑟𝑟𝑡𝑡𝑂𝑂𝑡𝑡 𝑆𝑆𝑟𝑟𝐿𝐿𝑖𝑖𝑡𝑡 (𝑁𝑁𝑁𝑁𝑁𝑁)=𝐵𝐵𝑡𝑡𝑂𝑂𝑡𝑡𝑓𝑓𝐿𝐿𝑡𝑡𝑟𝑟−𝐶𝐶𝑠𝑠𝑟𝑟𝑡𝑡𝑟𝑟
As a way o unde s anding he cos s and bene i s, a cos -bene i balance shee may be de eloped. The
cos -bene i balance shee o he cons uc ion o he second na ional ai po in he Ne he lands (in
supplemen o exis ing na ional ai po a Schiphol) is p esen ed below (Wesemann, 2000) in Table 3.12.
Cos s
Bene i s
Cons uc ion cos s
Ope a ing e enue
Modi ica ion o ai space s uc u e
Ne e enue om passenge s and eigh
O he cos s (including oad a ic in as uc u e)
Indi ec economic e ec s
Noise nuisance a new ai po
Noise nuisance a Schiphol-
Planning assimila ion
Employmen oppo uni y
O he e ec s
Table 3-11 Cos -bene i balance shee o a second na ional ai po in he Ne he lands (Wesemann, 2000)
phoebe-p ojec .eu
Copy igh © by PHOEBE
85
Deli e able 1.1 – S a e o he A and end-use needs e iew
The analysis mus ca y ou p ecise calcula ions o he RTA cos s and bene i s o calcula ing he CBR and
NPV. The e exis mul iple me hods o calcula e RTA cos s. Figu e 3-14 p esen s he di e en me hods o
calcula ing RTA cos s (Baum e al., 2000).
The e a e se e al me ics o calcula ing o es ima ing he bene i s o oad sa e y measu es. One such
me ic is he alue o s a is ical li e (VOSL). VOSL is also called he alue o li e, he alue o p e en ing
a ali y (VPF) o he implied cos o a e ing a a ali y (ICAF) (Johansson, 2001; Vecino-O iz & Hyde ,
2014).
Figu e 3-14 Me hods o calcula ing RTA cos s (Baum e al., 2000)
One example o cos bene i analyses o oad sa e y assessmen is he iRAP Sa e Roads In es men
Plans (SRIP). The SRIP is one o he ou pu s o an iRAP assessmen and aim o guide u u e oad ne wo k
sa e y upg ades. I p o ides a p io i ised lis o coun e measu es ha can cos -e ec i ely educe
in as uc u e- ela ed isk. Compa ing he cos o implemen ing he coun e measu e wi h he educ ion in
c ash cos s esul ing om i s implemen a ion, he model can iden i y he sa e y ea men s ha a e
economically ad an ageous o implemen o p io i ise in es men s. Some esea ch al eady showed ha o
e e y inc ease in he s a a ings he cos s o he c ashes pe km is almos hal ed (Figu e 3-) (McIne ney
and Fle che , 2013).
The bene i s a e calcula ed based on he o al es ima ed numbe o a ali ies and se ious inju ies (FSI) on
he ne wo k ha he implemen a ion o he coun e measu e ecommended can sa e. FSI es ima es a e
made o each 100m segmen o he exis ing oad unde exis ing condi ions. Whe e S a Ra ings ep esen
he indi idual isk o a oad o a use (i.e. likelihood and se e i y o a c ash), a ali y es ima ions ep esen
phoebe-p ojec .eu
Copy igh © by PHOEBE
86
Deli e able 1.1 – S a e o he A and end-use needs e iew
collec i e isk which accoun s o exposu e (i.e. oad use olumes) and ac ual c ash a es. Mo e
in o ma ion abou he FSI calcula ion can be ound in he iRAP Me hodology Fac shee n10 – Casual y
Es ima ion and Calib a ion. (iRAP, 2014). The economic alue o he bene i s a e ep esen ed by he
Human Capi al da a, ins ead o WTP da a o calcula e VOSL (Vecino-O iz & Hyde , 2014). iRAP p o ides
a se o equa ions ha quan i y he VOSL as a unc ion o G oss Domes ic P oduc (GDP) pe capi a
(Vecino-O iz & Hyde , 2014 ).
Figu e 3-15 The ela ionship be ween S a Ra ings and he cos o a ali ies and se ious inju ies
The iRAP model can analyse mo e han 120 oad imp o emen op ions. They ep esen commonly used
enginee ing ea men s ha a e on he lis suppo ed by hei e idence on educing oad sa e y isk. Mo e
in o ma ion abou he sa e y ea men s conside ed in he model can be ound in he iRAP Me hodology
Fac shee n11. (iRAP, 2023). Some sa e y ea men s ha e been de eloped o explici ly ul il he need o
incen i e sa e speeds in u ban a eas and p omo e sa e a els o VRUs.
3.4.2 Cos -E ec i eness Analysis
The Cos -E ec i eness Analysis (CEA) is ano he me hod o analyse he socio-economic impac o oad
sa e y measu es. Me hodologically, CEA is closely ela ed o CBA. In some ins ances, CEA is conside ed
a a ian o CBA (Wesemann, 2000). In CEA, we weigh he cos s o oad sa e y measu es agains hei
e ec s. In his case, he e ec s a e no exp essed mone a ily (Baum e al., 2000; OECD, 2001). One way
o ca ying ou CEA o e alua e a pa icula oad sa e y measu e is by using Inc emen al Cos E ec i eness
Ra io (ICER) (Chan i h e al., 2021). ICER is also used o ank oad sa e y measu es in e ms o hei
e ec i eness. Ma hema ically, ICER is desc ibed as:
𝐼𝐼𝐶𝐶𝐸𝐸𝑆𝑆 = 𝐴𝐴𝑆𝑆𝑡𝑡𝑟𝑟𝑟𝑟𝑂𝑂𝑡𝑡 𝐷𝐷𝐿𝐿𝑓𝑓𝑓𝑓𝑡𝑡𝑟𝑟𝑡𝑡𝑂𝑂𝑡𝑡 𝑠𝑠𝑓𝑓 𝐶𝐶𝑠𝑠𝑟𝑟𝑡𝑡 (𝐴𝐴𝐷𝐷𝐶𝐶)
𝐴𝐴𝑆𝑆𝑡𝑡𝑟𝑟𝑟𝑟𝑂𝑂𝑡𝑡 𝐷𝐷𝐿𝐿𝑓𝑓𝑓𝑓𝑡𝑡𝑟𝑟𝑡𝑡𝑂𝑂𝑡𝑡 𝑠𝑠𝑓𝑓 𝑖𝑖𝑂𝑂𝐿𝐿𝑡𝑡 𝑠𝑠𝑓𝑓 𝐸𝐸𝑓𝑓𝑓𝑓𝑡𝑡𝑠𝑠𝑡𝑡𝐿𝐿𝑆𝑆𝑡𝑡𝑂𝑂𝑡𝑡𝑟𝑟𝑟𝑟 (𝐴𝐴𝐷𝐷𝐸𝐸)
ADC ep esen s he a e age change in he budge in he gi en ime in e al ( he pe iod om no oad sa e y
measu e ill i s engagemen ). In con as , ADE ep esen s he a e age change in he numbe o dea hs
caused by RTA du ing he same ime in e al. The esul an ICER o each measu e is hen accoun ed o
along wi h he ou quad an s o he CEA plane shown below in Figu e 3-16 and Figu e 3-17 (Chan i h e
al., 2021).
phoebe-p ojec .eu
Copy igh © by PHOEBE
87
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 3-16 CEA plane ( ou quad an ) (Chan i h e al., 2021)
Figu e 3-17 Quad an explana ion (Chan i h e al., 2021)
3.4.2.1 Willingness o Pay me hod (WTP)
To explain he WTP me hod, a scena io is conside ed. Suppose a oad sa e y measu e o policy is expec ed
o educe he numbe o dea hs caused by RTAs by 𝒏𝒏 o e a gi en ime ame o a popula ion o size 𝑷𝑷.
Now, i 𝒗𝒗 is he a e age amoun each membe o he popula ion is willing o pay o und he oad sa e y
measu e o isk educ ion, hen he o al amoun o be paid by he popula ion is 𝒗𝒗.𝑷𝑷. Thus, he willingness
o pay pe a ali y is 𝒗𝒗.𝑷𝑷/𝒏𝒏. Acco ding o he WTP me hod, his is also known as he alue o p e en ing a
a ali y (VPF) (OECD, 2001).
Two di e en empi ical app oaches can es ima e WTP o a gi en popula ion:
• Re ealed p e e ence me hod (OECD, 2001)
• S a ed p e e ence me hod (also known as con ingen alua ion) (An oniou, 2014; Rizzi e al., 2006).
3.4.2.2 Mul i-C i e ia P ocesses
As he name sugges s, MCP elies on a ious explici c i e ia o assessmen . These c i e ia can di e
signi ican ly. Wi hin MCP, sco es a e decla ed o e e y indi idual c i e ion. These sco es ha e hei
espec i e indi idual uni s and hus canno be agg ega ed o e all he c i e ia. Mo e signi icance can be
a ibu ed o speci ic c i e ia in MCP while conduc ing he assessmen . The deg ee o impo ance o he
c i e ia can be egula ed by assigning sui able weigh s o he c i e ia. In case he e is a signi ican
di e gence in he opinion amongs he decision-make s o policymake s, mul iple se s o weigh ing ac o s
can be added. Finally, he c i e ia a e anked acco ding o hei impo ance (weigh ing ac o s) (OECD,
2001; Wesemann, 2000). An example o MCP is gi en below in Figu e 3-18.
phoebe-p ojec .eu
Copy igh © by PHOEBE
88
Deli e able 1.1 – S a e o he A and end-use needs e iew
A ew MCP me hods a e lis ed below(OECD, 2001; Wesemann, 2000):
• Weigh ed agg ega ion me hod
• Goals achie emen ma ix
• Conco dance analyses,
• Pe mu a ion me hod
• Regime me hod
• Mul i-dimensional scale analysis
• E amix app oach.
Figu e 3-18 Example o an MCP (Kanugan i e al., 2017)
phoebe-p ojec .eu
Copy igh © by PHOEBE
95
Deli e able 1.1 – S a e o he A and end-use needs e iew
li e a u e. The e iew s udy conduc ed by Biljecki & I o (2021) showed ha his ype o da a has been used
in oad sa e y e alua ions o conduc i ual s ee audi s, explo e ela ionships be ween c ashes and oad
ea u es, and unde s and he pe cep ion o sa e y. O he au ho s ha e ocused on iden i ying lows and
speeds using image ecogni ion, which is pa icula ly impo an o oad sa e y (Yin e al., 2015). The Wo ld
Bank ein o ces he po en ial o image ies o complemen o eplace adi ional da a collec ion o oad
sa e y assessmen s (Wo ld Bank, 2021).
Many challenges exis in de ec ing a ibu es h ough image ecogni ion (Sanjeewani & Ve ma, 2019). The
accu acy o he esul s is highly suscep ible o he quali y o he images cap u ed. Since mos sys em
seman ics segmen he pixels, i 's essen ial o use he igh equipmen o minimise he impac o na u al
ligh ing (To baghan e al., 2022). Mo eo e , he algo i hms need o be ained, which in ol es, once again,
manual inpu s and he esul s a e highly sensi i e o he numbe o aining i e a ions ca ied ou
(Sanjeewani & Ve ma, 2019). Finally, s udies epo he di icul y coding speci ic a ibu es such as ees
mixed in ege a ion and hin objec s (Lucchesi e al., 2023; Ve ma e al., 2018). The s udies conduc ed by
Sanjeewani & Ve ma (2019) and Ve ma e al. (2018) ha aimed o de ec a ibu es o he AUSRap model
au oma ically ob ained high le els o accu acy o bicycle pa hs and some ypes o e ical signs (e.g.,
speed and cu a u e signs) bu e y low o o he a ibu es, including pa emen de ec s and he p esence
o umble s ips. The au ho s sugges ed hei esul s a e p omising bu ha mo e wo k is needed.
Ne e heless, image da abeses a e no always accessible o academic and p ac i ione s. Al hough he e
a e cu en ly many da a p o ide s, da a is no easily licensed. Besides, ex ac ing a ibu es om images
usually con lic s wi h he e ms and condi ions o hese pla o ms (Biljecki & I o, 2021) ha ha e con inuously
inc eased use es ic ions (Fang e al., 2021). C owdsou ced pla o ms aim o o e come hese challenges
collabo a i ely (e.g, Mapilla y), bu he lack o s anda d and quali y assu ance inc ease p ocessing ime and
educe applicabili y. Also, p o ide s end no o p o ide he measu emen pa ame e s equi ed o co ec
image dis o ions. Thus, geome ic dimensions mus be used wi h cau ion (Yin e al., 2015).
The second mos common app oach o image ecogni ion is acking mo emen s using oo age o s ee s
and oads, especially om oad su eillance came as (Y. Wang e al., 2022). P esen ing a g owing use in
beha iou al s udies, ideo acking can p o ide in o ma ion ega ding speed, accele a ion, angle o
swe ing, oad posi ion, ollow-up pa ame e s and o he s (Sohail e al., 2023; To baghan e al., 2022).
Using ideo da a and ollowing he co esponding ideo p ocessing echniques, he iden i ica ion o a ic
con lic s is applicable as ha e been done in many s udies aiming o alida ing con lic s ex ac ed h ough
a ic simula ion (As a i a e al., 2012; Li e al., 2016; Essa and Sayed, 2020) and o in es iga e
ans e abili y o model pa ame e s (Essa and Sayed, 2015).Al hough he e iew pape s ound in his
li e a u e e iew p esen mos ly ideo acking o ehicles, he sys ema ic sea ch did ind some
expe imen s wi h ulne able oad use s. G uden e al. (2019) used eco ding om a pedes ian amp o
es ima e and model pedes ian lows. An indus y example came om Vi aci y. The o ganiza ion acked
pedes ian mo emen s o unde s and changes in beha iou and use o he space a e in as uc u e
changes implemen ed in Hackney, London a Fai hol Road and Rendlesham Road (Vi aci y, 2023).
Su eillance came a da a is a good i o longi udinal s udies, cap u ing in o ma ion om di e en pe iods
and days o he week. Howe e , geog aphical co e age is limi ed (To baghan e al., 2022). On he o he
hand, s udies epo high accu acy and eliabili y o pa ame e s ex ac ed om ideo acking me hods
(Gu ie ez-Oso io & Ped aza, 2020).
3.5.1.4 In e ne -o -Things (IoT)
Much o he li e a u e conce ning he applica ion o he in e ne -o - hings o anspo is ocused on he use
o sma phones, which we discuss in he nex sec ion. Taha (2018) p oposes an IoT a chi ec u e o
assessing oad sa e y by inco po a ing oad condi ion me ics, using da a ob ained om wi hin ehicles in
andem wi h in o ma ion om o he senso sys ems. The au ho s we e able o plan a ou e such as o
minimise he exposed isk by combining da a om a ious li e sou ces in o a dynamic oad sa e y
phoebe-p ojec .eu
Copy igh © by PHOEBE
96
Deli e able 1.1 – S a e o he A and end-use needs e iew
assessmen . Senso s inside he ehicle included s and-alone accele ome e s, ca bon monoxide senso s,
and OBD-II de ices ha ex ac da a di ec ly om he con olle a ea ne wo k (CAN).
3.5.1.5 Sma phone as a senso
Sma phones con aining mul iple se s o senso s a e now ubiqui ous in many coun ies. These sma phones
ypically con ain GNSS posi ioning sys ems, accele ome e s, magne ome e s, mic ophones, and came as.
The la ge numbe o sma phones in ci cula ion means ha senso s a e egula ly passing o e a signi ican
ac ion o he oad ne wo k in some e i o ies. The widesp ead co e age p esen s a an alising oppo uni y
o unde s anding he anspo ne wo ks in unp eceden ed de ail. The size o he da ase s p esen s i s own
p oblem in e ms o inco po a ion in o p ojec s by anspo and oad sa e y p o essionals. GNSS da a a e
usually sampled a a maximum a e o 1 Hz, whe eas accele ome e da a can be cap u ed a equencies
10 − 100 Hz. Sma phone moni o ing o GNSS da a o a jou ney las ing en minu es and sampled a 1Hz
will c ea e a da ase o 600 en ies. Reco ding high- equency accele ome e da a will c ea e a da ase o
60,000 en ies in he same ime pe iod. Conside ha s udies o p ojec s may moni o housands o
indi iduals o hund eds o housands o ips each, and he olume o da a a ailable can easily swell o
ens o billions o en ies. Needless o say, such da a equi e signi ican amoun s o agg ega ion, o en
coupled wi h ML, in o de o p oduce ac ionable insigh s.
Many esea che s ha e in es iga ed he use o sma phones o oad su ace moni o ing by in es iga ing
he ex en o which a e sals o e damaged su aces p oduce a cha ac e is ic signal in accele ome e da a
(Mednis e al., 2012; Bello e al., 2019). Recen ly, Ma a azzo e al. (2022) demons a ed ha da a om
sma phone accele ome e s can be used o moni o he modal equency o b idges; po en ially p o iding a
cheape al e na i e o adi ional me hods o moni o ing in as uc u e.
Mo e esea ch has a emp ed o unde s and human beha iou using mobile elephones wi h GNSS
capabili ies, which ac ually p eda es he cu en e a o sma phones (e.g. Gonzalez e al., 2008). Such
beha iou s ha e included hose o he d i e , such as he e ec s o a igue (Chan e al., 2019) and he
ela ionship be ween di e en accele a ion beha iou s and he p opensi y o c ash (a Wåhlbe g, 2008; a
Wåhlbe g, 2012). Many s udies o oad use s concen a e on examining he ex en o which he use is
dis ac ed by he sma phone. The e is al eady a subs an ial ma ke in he insu ance sec o o d i e
moni o ing se ices ha u ilise in-ca de ices, he majo i y o which will be sma phones. These p i a e
companies, such as The Floow, use sma phone da a o unde s and how indi idual d i e s compa e o
es ablished s anda ds in e ms o bo h he ehicle dynamics (speed, accele a ion e en s, ha sh b aking)
and human ac o s such as a igue and he choice o a el ime. The widesp ead use o elema ics has
impo an implica ions o e i o ies whe e mo o insu ance is a manda o y p e equisi e o he legal
ope a ion o a ehicle; c ea ing a la ge sample o moni o ed d i e s and equen sampling o he oad
ne wo k. The da a collec ed can be agg ega ed and, in u n, used o c ea e no el solu ions o es ablished
and eme ging mobili y- ela ed p oblems. The e o e, as a consequence o p o iding widesp ead moni o ing
o insu ance pu poses, The Floow can in- u n p o ide s a is ical measu es o speed and ho ough a e a ic
o oad sa e y and local go e nmen s akeholde s (h ps://www. he loow.com/mobili yIn/).
The e a e nume ous s udies as well as na u alis ic d i ing p ojec s (e.g. he i-DREAMS Ho izon 2020
p ojec ) ha moni o ac i e a el ips using sma phones, howe e , such s udies end o be ocused on
he de elopmen o an IoT sys em ha would allow ehicles o p edic he passage o pedes ians on o he
oad (such as Ridel e al., 2018). While some s udies do collec GNSS aces om he sma phones o
cyclis s (e.g. Rupi & Schweize , 2018), he e is li le wo k in o using such da a o c ea e ac ionable insigh s
o anspo p o essionals.
3.5.1.6 Geog aphic In o ma ion Sys em (GIS)
Ci ies ha e al eady ecognised he impo ance o ha ing solid GIS da abases o u ban ea u es. Combining
di e en laye s o in o ma ion in o a spa ial da a s uc u e pe mi s easy o isualise and analyse any spa ial
dependen da a (Rez ani e al., 2023). Due o olun ee ed Geog aphic In o ma ion (VGI) pla o ms, GIS
phoebe-p ojec .eu
Copy igh © by PHOEBE
97
Deli e able 1.1 – S a e o he A and end-use needs e iew
da a ha e sp ead among de elope s, indus y ac o s, esea che s, and o he end use s. The mos well-
es ablished VGI pla o m is Open S ee Map (OSM). OSM ope a es collabo a i ely among use s ha map
hei su oundings and upda e he in o ma ion o he global pla o m. OSM also cap u es in o ma ion on
publicly a ailable GIS da a o o icial go e nmen en i ies and accep s a con ibu ion o indi idual GPS
uni s (Foody e al., 2017).
C owded-sou ces GIS pla o ms dissemina e spa ial da a like ne e be o e. The OSM was de eloped o
ul il a la en need o eely a ailable map da a; he mos well-known mapping sys em is copy igh ed and
da a canno be ex ac ed o academic o comme cial use. Howe e , he e a e h ee in insic limi a ions o
his ype o p ojec . The i s is ha mo e da a quali y assu ance is need (Goodchild & Li, 2012), as use s
a e no equi ed o p o e expe ience o skills in mapping, which means he e needs o be alida ion be o e
he da a a e made a ailable o o he use s (Teimoo y e al., 2021). Secondly, he e a e no s anda ds o
da a upda es. While he mos ecen da a a e a ailable, he use will ha e access o loca ional da a ha is
empo ally incompa ible. Mo eo e , some au ho s indica e ha he numbe o e sions, use s, and he
empo al a ia ion ange a e ele an o da a eliabili y (Teimoo y e al., 2021). Finally, OSM da a p esen
he e ogeneous accu acy o di e en egions (Foglia oni & Kauppinen, 2014). S udies ha used o icial
go e nmen da a, when sui able and a ailable, o compa e wi h OSM ag ee ha da a is o highe quali y
and a ailabili y a e in u banised a eas (Minghini & F assinelli, 2019).
Besides he conce ns wi h da a quali y, he OSM has been widely used as a ounda ion o oad sa e y
analysis as a da a sou ce o ne wo k iden i ica ion, oad ea u es and oad ype. The Wo ld Bank (2021b)
epo De ec ing U ban Clues o Road Sa e y Le e aging Big Da a and Machine Lea ning ein o ced he
impo ance o using OSM da a in combina ion wi h o he c owd-sou ced da a sou ces like oad ale s om
Waze and image a ibu e de ec ion based on Mapilla y image y.
The combina ion o di e en da a sou ces b ings a key elemen o discussion: he abili y o GIS pla o ms
o pe o m da a usion. A ypical issue academics and p ac i ione s ace geospa ial da a con la ion when
di e en da a sou ces p o ide in o ma ion o he same ea u e bu need he exac coo dina es (Khan e al.,
2010). The li e a u e b ings examples o many map-ma ching algo i hms. They a e had been dependen
upon he cha ac e is ics o he da a elemen s, he le el o accu acy equi ed, he le el o complexi y o he
oad ne wo k, and he e en ual use o he da a. Combining GIS da a wi h o he da a ypes, like
socioeconomic da a and su eys, is also common in s udies wi h ulne able oad use s. Li e a u e e iew
s udies poin ed o his app oach as widely used o co ela e cycling beha iou wi h buil en i onmen
measu es (Yang e al., 2019).
Mo eo e , GIS pla o ms ha e been epo ed o acili a e mul ic i e ia analysis and geos a is ical da a
ea men (Rueda-Villa e al., 2019). The e is well-es ablished knowledge o i s po en ial o suppo c ash
clus e ing and iden i ica ion o ho spo s (Aghasi, 2019). GIS also plays an impo an ole in da a
isualiza ion. Geospa ial da a isualiza ion can bene i decision-make s and non-expe in unde s anding
oad sa e y issues and ad oca ing o change (Sanguine i & Als on-S epni z,2023).
3.5.1.7 Ae ial Vehicles
Manned Ae ial Vehicles (MAVs) enable a bi a y an age poin s (con olled by egula ed lying es ic ions),
which can make hem a ac i e o da a ga he ing ac oss wide and emo e geog aphy and a loca ions no
sui ed o g ound based o o he moni o ing. Despi e hese ad an ages MAVs emain p ohibi i ely expensi e
wi h e y high ope a ing cos s. Howe e , MAVs ha e a niche oppo unis ic applica ion owa ds oad sa e y
when coupled wi h o he uses such as suppo ing eme gency a ic managemen , law en o cemen
ac i i ies, inciden esponse, and spo adic ae ial su ey asks6. Da a om manned ae ial ehicles is a ely
6 Fo ins ance in pa s o Aus alia and some US s a es whe e oad sa e y and a ic asks may upon
occasion u ilise manned ae ial ehicles also used o o he pu poses.
phoebe-p ojec .eu
Copy igh © by PHOEBE
98
Deli e able 1.1 – S a e o he A and end-use needs e iew
a ailable wi hou high cos s and ypically has e y limi ed empo al and spa ial co e age making usage
imp ac ical o he PHOEBE p ojec and mos o he oad sa e y usage.
Unmanned Ae ial Vehicles (UAVs), o en e e ed o as ‘d ones’, o e signi ican ad an ages o e MAVs,
wi h bo h lowe pu chase and ope a ional cos s. UAVs o en ha e supe io manoeu abili y, which enables
p ecise posi ioned ae ial da a ga he ing (Ou ay e al, 2020). UAVs can be equipped wi h a wide ange o
emo e sensing equipmen (such as ligh weigh UAV LIDAR o isible spec um came as). Howe e , hese
capabili ies a e limi ed by a ious ac o s:
1) Weigh - senso s ,da a p ocessing and ansmission ha dwa e and ba e y capaci y limi po en ial
ligh ime (cu ailing da a ga he ing o ypically much less han 1 hou ).
2) Regula ions - UAVs mus ope a e unde ligh ceilings, es ic ions nea ai ields, es ic ions o e
u ban a eas, es ic ions on nigh ime lying and needs o licensed ope a o s which challenges
da a ga he ing minimising he po en ial o new low cos app oaches7.
UAVs can be u ilised in a ange o oad sa e y a eas such as
3) Ex ac ing ehicle ajec o ies om imaging by emo ely ga he ing snapsho samples o loca ional
isk p oxy da a, such as he ime o collision. Such da a can also be ha es ed om connec ed
ehicles and la ge scale ‘nea miss’ da a wi h highe spa ial and empo al co e age han ae ial
me hods (Khan e al, 2017).
4) T a ic inciden scene da a cap u e, such as cap u ing LIDAR da a a ound inciden si es o la e
o ensic analysis. Despi e his capabili y, ae ial-based LIDAR o e s only ma ginal ad an age o e
g ound-based LIDAR o oad based inciden s (Mehmood e al, 2018).
5) Speed en o cemen and moni o ing - (Ba mpounakis & Ge oliminis, 2020)
6) Fea u e ex ac ion along ou es o in e sec ions o unde s and a ibu es ele an o sa e y (Chen
e al., 2017 and As o , 2023). Such da a can also be ha es ed om g ound-based ehicles o
su eys a a lowe cos .
Despi e he po en ial ad an ages om ae ial app oaches, he need o ce i ied ope a o s and he egula o y
es ic ions upon use in popula ed egions p e en s adequa e usage in u ban se ings as may be equi ed
in PHOEBE.
3.5.1.8 Social Media
Ci izen social media ac i i y can p o ide a po en ial sou ce o opinions and sha ed in o ma ion ela ed o
mobili y and i s sa e y. A la ge numbe o s udies ha e explo ed he use o na u al language p ocessing
echniques and da a mining8 om la ge scale social media o suppo a a ie y o anspo analysis asks.
T anspo sa e y ela ed asks include:
1) Iden i ying he ime, loca ion and se e i y o oad inciden s whe e o he da a sou ces emain
incomple e o supplemen and add o oad sa e y inciden e idence (Saelm e al, 2021).
2) C owdsou cing public opinion ela ed o oad sa e y. This may be speci ic and a ge ed o localised
issues o hema ic sa e y conce ns ac oss socie y. (Zaye e al, 2021).
7 UK egula ions (seen as an exempla in d one egula ion) con ain speci ic guidance o use o d ones in
ela ion o oads, a ic managemen and oad Sa e y. Simila guidance (albei less ocused upon oad
sa e y usage) is p o ided in he EU, US, China as well as o he in e na ional coun ies.
h ps://www.s anda ds o highways.co.uk/dm b/sea ch/ d 64db4-6745-4d3d-ae18-62dad0ae1225
8 P incipal analysis u ilises, sen imen analysis, geoloca ion NLP analysis and empo al analysis o y and
de ine usable da a om uns uc u ed con en .
phoebe-p ojec .eu
Copy igh © by PHOEBE
99
Deli e able 1.1 – S a e o he A and end-use needs e iew
3) Iden i ying ne wo k s a e by iden i ying eal ime a ic in o ma ion ha may be sha ed o mally o
in o mally ia social media channels. (Ade iloye & Awas hi, 2018).
Despi e he abo e aims he use o social media has se e al undamen al challenges:
1) Da a p o ec ion and pe mi ed p ocessing - This is challenging and can p esen e hical and da a
p o ec ion conce ns in he e-use o hi d pa y uns uc u ed da a ha equen ly con ains pe sonal
iden i iable in o ma ion. Al hough legi ima e in e es can p o ide consen when p ocessing can help
public bene i , p ocessing con ols and impac assessmen s need e y ca e ul conside a ion.
2) Da a bias - social media is undamen ally biased e en in oad sa e y usage whe e da a ex ac ed
ypically has demog aphic, language and communi y biases meaning i s ou pu should be u ilised
wi h s ong cau ion whe e used in decision making, moni o ing and analysis.
3) Geoloca ion challenges messages a e pos ed globally bu o en wi hou exac geoloca ion making
eliable euse o da a ha d o main ain. Mos social media channels do no con ain geoloca ion (o
a e minimising i s usage9) meaning ha o loca e he meaning o messages he loca ion mus be
in e ed om ex in he message, a p ocess wi h dis inc accu acy conce ns gi en non unique
geospa ial loca ion naming con en ions.
4) Tempo al challenges - messages equen ly e e o pas e en s o he po en ial o u u e e en s
making i ha d o de e mine despi e message imes amps when and he du a ion o a messages
subjec wi h p ecision and accu acy.
5) Subjec challenges - social media messages will use a iable desc ip o s and language o opics
discussed making i ha d o classi y and clus e e en s whe e sepa a ed desc ip ions may e e o
one o mo e e en s and de e mina ion o singula o mul iple occu ences is o en ha d o cla i y.
E en wi h such challenges social media emains a use ul ool o oad sa e y bu due o i s conce ns o
accu acy, bias and da a p o ec ion (Zaye e al, 2021) i may s ill p o ide some insigh on ci izen opinion o
PHOEBE ac i i y. Social Media howe e bes suppo s dissemina ion ac i i ies a he han obus e iden ial
da a ga he ing.
3.5.2 Summa y
Da a ga he ing and colla ion echniques ac oss li e a u e (and in p ac ice) a e a ied and can be seen o
co e a numbe o base da a ga he ing equi emen s:
1) Physical a ibu e da a - de ailing he loca ional a ibu es o an en i onmen ha may be co ela ed
wi h i s sa e y.
2) Loca ion ocused use beha iou da a - de ailing he loca ional use o he en i onmen ha may be
co ela ed wi h i s sa e y.
3) Gene al use da a - de ailing he backg ound o use s as may in luence isk unde s anding and
pa e ns o beha iou .
4) Ac ual inciden da a - o indica e whe e nega i e ou comes ha e p e iously occu ed o unde s and
causa ion and s a is ical co ela ions o hem.
Wi hin each a numbe o sub da a needs a e de ailed in li e a u e whe e ound o be suppo i e o oad
sa e y unde s anding.
Also as well as da a needs li e a u e de ails a numbe o me hodological app oaches o ob ain o imp o e
a ailable da a o oad sa e y. These again consis o wide me hodologies o da a ga he ing ha can be
colla ed in o dis inc ypes o da a ga he ing me hodologies. These a e:
9 Fo ins ance Twi e is in he p ocess o emo ing p ecise geoloca ion om i s eco ded social media pos s
o help suppo da a p o ec ion and align o global da a p o ec ion good p ac ices.
phoebe-p ojec .eu
Copy igh © by PHOEBE
100
Deli e able 1.1 – S a e o he A and end-use needs e iew
1) Indi idual colla ed da a - manual and human collec ed in o ma ion ga he ing o i s analysis and
applica ion. Da a ga he ed ia di ec inspec ion, manual me hods o di ec consul a ion wi h
s akeholde s o public opinions. Such da a ga he ing is o en ime consuming and cos ly so is
ypically only pe o med when no al e na i e is a ailable and he need jus i ies he di ec e o s
equi ed.
2) T a el ou e emo e sensing and moni o ing - mobile sensing om ehicles o along a anspo
ou e o de e mine cha ac e is ics along in as uc u e. Me hodologies include de ice based means
o ga he da a and analysis o applica ion o oad sa e y.
3) O e head and ne wo k da a - ae ial, sa elli e, mapping o anspo ne wo k da a when u ilised o
ob ain ea u es o insigh o oad sa e y.
4) Fixed poin mobili y sensing - in as uc u e ixed poin senso s obse ing mobili y and beha iou s.
Me hodologies use a ange o ha dwa e i ed in p oximi y o mobili y in as uc u e coupled wi h
analysis echniques o gain insigh ela ed o sa e y.
5) Inciden da a - he ga he ing and euse o p io inciden eco ds, s a is ics and nea miss da a.
App oaches use a ious me hods o colla e da a ela ed o loca ional inciden s o isk p oxies o
hem (e.g. nea miss da a).
6) Thi d pa y da a euse - oad sa e y suppo ing da a such as wea he o icke ing da a ha may be
eu ilised when suppo ing oad sa e y analysis.
Ac oss he needs and me hodologies in li e a u e some me hodologies a e commonly epo ed o sol e a
pa icula need, whils o he s a e me ely suppo i e o he da a need and may be used alongside o he
app oaches. In many cases me hodologies a e only sui ed o speci ic needs, hese a e shown in he
ollowing ma ix in Figu e 3-20.
3.5 Discussion on e iew indings
The de elopmen o he PHOEBE F amewo k in ol es h ee p incipal aspec s:
1. Fi s is he heo e ical de elopmen , which includes he esea ch, whe e necessa y, o key p inciples
unde pinning he amewo k, and he me hodology o he amewo k
2. Second is he echnical de elopmen s equi ed o ensu e ha he echnical componen s o he
amewo k ha e he necessa y unc ionali y o mee he amewo k’s needs and p oduced he analysis
equi ed, and
3. Thi d a e he ools and suppo ma e ials o end use s o ensu e he amewo k can be u ilised o i s
s a ed pu pose.
The pu pose o his e iew was o documen he cu en s a e-o - he-a o each componen o he PHOEBE
F amewo k and highligh any pa icula conside a ions which should be aken in o accoun in i s
de elopmen .
phoebe-p ojec .eu
Copy igh © by PHOEBE
101
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 3-20 Da a needs e sus iden i ied da a ga he ing me hodologies
3.5.1 T a ic simula ion, human beha iou and oad sa e y assessmen
The PHOEBE F amewo k aims o ad ance he applica ion o a ic simula ion ools and oad sa e y
assessmen o enable anspo planne s and manage s o ully unde s and and add ess he sa e y
implica ions o changes in oad condi ions, mode choice, new modes, oad use beha iou s and o he
ac o s and hei impac s on sa e y o e ime.
T a ic mic osimula ion is he name o he echnique known o ep oducing he beha iou o each single
ehicle in he a ic low. This is done by simula ing each ehicle ajec o y wi h a high le el o de ail, his
includes, speed accele a ion, cu en lane, ou e, e c. This means ha mic osimula ion is e y complex,
being many di e en models in ol ed in o he simula ion.
phoebe-p ojec .eu
Copy igh © by PHOEBE
102
Deli e able 1.1 – S a e o he A and end-use needs e iew
Fo he PHOEBE p ojec , he ocus will be on: (1) he ehicle dynamics wi hin he lane such as ca - ollowing
beha iou s, bu also be ween lanes such as lane-changing beha iou s and how sa e/unsa e beha iou is
inco po a ed in hese models; (2) how ulne able oad use s a e cu en ly modelled and conside ed in
mic osimula ion ools and hei in e ac ion wi h o he non-mo o ized and mo o ized a ic a ne wo k le el;
and (3) ne wo k-le el sa e y assessmen app oaches ha can be used o imp o e sa e y analysis.
3.5.1.1 Mic oscopic a ic simula ion ools
T a ic mic osimula ion is ypically pe o med using one o many comme cial and open-sou ce so wa e
packages. The de elopmen o , and ma ke o , a ic simula ion models has been d i en in pas decades
by he need o design and op imise oads and oad ne wo ks wi h a iew o minimising conges ion and
maximising h oughpu .
While he e a e many di e ences among hem, when i comes o sa e y mos o hem beha e in a simila
manne . This is because ehicle dynamics, while ealis ic on an agg ega e le el, do no e lec he d i e ’s
beha iou al de ails a a mic oscopic (o disagg ega ed) le el.
Ca - ollowing and lane-changing models ha mainly go e n he longi udinal and la e al mo emen s o
ehicles do no accoun o human ac o s ha migh lead o di e en sa e y ou comes in eali y. This means
mos mic osimula ion models a e concei ed as inhe en ly ‘c ash ee’.
3.5.1.2 E alua ing sa e y using con lic -based me hodologies
Con lic -based me hodologies which use su oga e sa e y measu es (such as ime o collision and gap
accep ance) a e used o e alua e a ic sa e y in mic oscopic a ic simula ion ools based on he
compu a ion o sa e y me ics by using simula ion ou pu s. This app oach does no need o al e he
modelled beha iou o d i e s o conside human e o s – allowing ehicle c ashes – in he simula ion
en i onmen , which in pa simpli ies he p oblem.
One o he mos ex ensi ely used app oaches o e alua e sa e y impac is he combina ion o mic oscopic
simula ion ou pu s wi h he Su oga e Sa e y Assessmen Model (SSAM). This me hod uses ehicle
ajec o y da a om simula ion ou pu s o obse e ehicle- o- ehicle in e ac ions, iden i y a ic con lic s
and he e o e, assess sa e y impac s on a ne wo k le el.
Despi e he signi ican p og ess achie ed ela ed o con lic -based me hodologies, he e a e limi a ions due
o he absence o comple e and es ablished models o simula ing po en ial c ashes. T a ic con lic s a e an
impo an indica o o a po en ial c ash isk, bu no all con lic s esul in c ash occu ences. The e o e, c ash
es ima ion p o ides a mo e accu a e and eliable assessmen o oad sa e y compa ed o jus es ima ing
con lic s.
While hese echniques a e mainly used om mic osimula ion da a (e.g., SSAM), he e is no clea e idence
on how eliable he esul s a e when compa ed o eal ajec o ies. O he d awbacks o SSAM a e ha VRU
in e ac ion wi h o he mo o ized ehicles a e no included and he selec ion o h eshold alues o some
su oga e sa e y indica o s is subjec i e. Only a ew s udies ha e a emp ed o o e come hese issues
h ough p oposing al e na i e app oaches and hence u he esea ch is equi ed. Mos o he pas esea ch
modelling e o s needed be e e alua ion o he c ash da a used, as well as he co esponding modelling
me hods ollowed (e.g. limi ed sample size, unde - epo ing, un eliabili y a e no icing). Add essing his gap
could lead o he de elopmen o be e s a egies o mi iga ing c ash occu ence and consequences, bu
would be beyond he scope o he PHOEBE p ojec o add ess i in his way.
The e o e, o he bes o he e iewe s’ knowledge, he e a e a leas wo po en ial app oaches o o e come
hese limi a ions: (a) imp o e he cu en SSAM app oach by de eloping new me hods ha
inco po a e he ajec o y da a om all oad use ypes – no jus ca s – and in eg a e i wi hin a
simula ion ool in o de o iden i y po en ial con lic s be ween all use s; and (b) adap and imp o e
sa e y indica o s om exis ing sa e y assessmen me hodologies (e.g., S a Ra ings).
phoebe-p ojec .eu
Copy igh © by PHOEBE
103
Deli e able 1.1 – S a e o he A and end-use needs e iew
3.5.1.3 Mic o-simula ion o non- ehicula oad use mo emen s and in e ac ions
I is e y impo an ha he PHOEBE F amewo k can iden i y isks o ulne able oad use s. The
coexis ence and in e ac ion o mul iple oad use s (e.g., mo o ized, non-mo o ized a ic and pedes ians)
in he same oad space c ea es complex si ua ions ha ha e no been p ope ly add essed in cu en
simula ion models.
T adi ionally, a ic simula ion so wa e packages ha e ocused p ima ily on ehicula a ic and hea y
ehicles, and ha e gi en li le o no a en ion o non-mo o ized oad use s. This is changing owa ds a mo e
in e ac i e simula ion en i onmen whe e all oad use s sha e space and eac o each o he .
The Ho izon 2020 P ojec MOMENTUM de eloped an ini ia i e o enhance a ic models o suppo
sus ainable u ban mobili y planning. The p ojec simula es scena ios o s udy inno a ions ela ed o
bicyclis s (e.g., bike-sha ing) and o he mic o-mobili y ends o help policymake s plan, manage and
egula e new anspo op ions (NOMMON, 2019)10. Howe e , he sa e y componen o his new mobili y
echnology o VRU is s ill no modelled and gene al models o unde s anding he in e ac ion be ween
pedes ians and non-mo o ised ehicle lows do no ye exis , despi e hese modes being cen al o
sus ainable anspo s a egies a ound he wo ld.
Si ua ions such as pedes ian c ossing lows and he esponse o d i e s a designa ed c ossings, and he
p esence o e-bikes and e-scoo e s and hei in e ac ion wi h o he oad use s a e some scena ios which
need o be conside ed in cu en simula ion ools o be e unde s and hei sa e y implica ions. Ex e nal
ac o s, such as wea he , cul u e, ime o day, si e con igu a ion, e c. can also ha e a s ong in luence on
oad use beha iou s.
3.5.1.4 Road sa e y assessmen
The e iew iden i ied a numbe o well-es ablished me hodologies which ep esen he cu en s a e-o - he-
a o how o e alua e in as uc u e- ela ed isk o a oad co ido o ne wo k. These included he S a
Ra ing me hodology which calcula es isk sco es o s anda d oad leng hs based on isk ac o s. The isk
ac o s used by he S a Ra ing me hod a e based on published C ash Modi ica ion Fac o s (CMFs) and
o he e idence in o oad c ash likelihood and se e i y. The S a Ra ings p o ide indi idual oad isk sco es
o VRUs (pedes ians, bicyclis s and mo o cyclis s) as well as ehicle occupan s.
The e iew also iden i ied se e al pu pose-speci ic and/o localised me hods, mos o which we e
de eloped using he S a Ra ings me hod o isk ac o s. These app oaches end o be mo e speci ic in
hei applica ion o he ypes o oads, oad use s, geog aphic egions, da a inpu s o analysis which a ec s
hei ans e abili y (e.g., high ocus on highway elemen s and simplis ic conside ing u ban elemen s) and
calib a ion (e.g., isk ac o s calib a ed o local c ash da a). Only a ew models ha e been p o en o be
applicable in b oade con ex s. Apa om he iRAP S a Ra ing and CycleRAP me hods, mos exis ing
models do no ocus on VRUs o u ban a eas despi e hei c i ical impo ance in u ban oad sa e y
assessmen .
This e iew con i med ha he S a Ra ings a e an app op ia e oad sa e y assessmen me hodology o
mee he needs o he p ojec aims. This me hod inco po a es indi idual models o VRUs, and is applicable
ac oss all egions and oad ypes (including u ban oad ne wo ks).
I is well-known ha ulne able oad use sa e y is highly dependen on he a ic and ehicle speed le els,
which in ci ies, can widely a y h oughou he day due o conges ion le els. All he me hodologies e iewed
in his epo do no ep esen he a iabili y o hese speed and low pa ame e s o e he cou se o a day.
10 NOMMON (2019). New Mobili y Op ions and U ban Mobili y: Challenges and Oppo uni ies o T anspo
Planning and Modelling. MOMENTUM - Ho izon 2020. URL: h ps://h2020-momen um.eu/wp-
con en /uploads/2020/01/MOMENTUM-D2.1-New-Mobili y-Op ions-and-U ban-Mobili y.pd
phoebe-p ojec .eu
Copy igh © by PHOEBE
104
Deli e able 1.1 – S a e o he A and end-use needs e iew
As ye , he S a Ra ing me hodology uses s a ic alues o a ic speeds and lows. I is planned unde he
PHOEBE p ojec ha a iable speed and low alues, in o med by a ic simula ion models, could
be used o p oduce dynamic isk a ings o di e en imes o he day. This could be used o in o m
be e FSI Es ima ion modelling o changes (and hus he socio-economic impac s), as well as help anspo
planne s unde s and and add ess peak isk imes.
3.5.1.5 Using human beha iou ac o s o calib a e mic oscopic a ic
simula ion
Human ac o s, such as cul u e, social in luence, a i udes owa d oad ule compliance, a igue, and isk-
compensa ing beha iou s, can all impac a ic sa e y ou comes. I human ac o s we e in oduced in
mic oscopic a ic simula ion models, iche beha iou al modelling de ail could be achie ed, and ehicle
c ashes would also be able o be modelled wi hin he simula ion.
While he li e a u e shows ha he in e ac ion be ween oad use beha iou and oad isk has po en ial o
u he ou unde s anding abou oad isk on oads, only a ew beha iou al ac o s such as isk pe cep ion,
e o s, and dis ac ion has been s udied in he pas , and o hose, mos ocussed on ca d i e s om he
pe spec i e o he indi idual’s beha iou .
Howe e , hese human ac o s a e di icul o inco po a e in cu en a ic simula ion models and mo e
de ailed da a – which is no o en a ailable – is needed o alida e and calib a e he models. Fu he mo e,
by adding mo e pa ame e s and ac o s inside models, i migh also ha e an impac on inc easing he
compu a ional cos , as i migh slow down simula ion unning imes.
I also aises he ques ion on which human ac o s should be p io i ised. Cu en ly, ypical de ec o da a
migh be su icien o calib a e adi ional a ic low-o ien ed models like Gipps (1981) and T eibe e al.
(2000) bu a e no enough o calib a e he le el o de ail equi ed o human ac o -based models. While
p og ess has been made in ecen yea s (Cal e e al., 2020), he e is no any model ha ul ils all he
equi emen s. Hence, decisions mus be made o decide which human ac o s a e mo e ele an o
he PHOEBE p ojec and wha da a is a ailable o calib a e models and alida e any new
de elopmen s.
3.5.1.6 Key conside a ions
The me hodological amewo k in PHOEBE should:
• Be de eloped conside ing no jus adi ional d i ing kinema ic ac o s, such as speed,
accele a ion, e c., bu also inco po a e o he human ac o s ha may lead o po en ial human
e o s when d i ing in simula ion models. In doing so, po en ial con lic s – and e en ehicle
c ashes – migh be able o be ep esen ed and quan i ied in a simula ion en i onmen .
• In es iga e which high- esolu ion human beha iou al da a could be collec ed h oughou he
p ojec , es ablishing a da a collec ion me hodology ha will be used o calib a e and alida e
a ic mic osimula ion models. I is also impo an o conside ha highly complex human ac o
models will signi ican ly inc ease he simula ion unning ime.
• Conside he imp o emen o cu en SSAM app oaches by de eloping new me hods ha
inco po a e he ajec o y da a om all oad use ypes – no jus ca s – and in eg a e i wi hin
a a ic mic osimula ion ool o iden i y po en ial con lic s be ween all use s.
• Build on he H2020 MOMENTUM p ojec ad ances o VRUs and sha ed mobili y simula ion
and he iRAP VRU a ali y and se ious inju y es ima ion models o es ablish a me hodological
p ocess o inco po a ing non-mo o ised/ligh anspo modes in a ic simula ion o p edic
oad sa e y ou comes o hese oad use g oups.
• U ilise exis ing sa e y indica o s, namely he iRAP S a Ra ings, o build on he cu en s a e-
o - he-a and expand hei applica ion in o new con ex s (in his case, mic oscopic a ic
phoebe-p ojec .eu
Copy igh © by PHOEBE
111
Deli e able 1.1 – S a e o he A and end-use needs e iew
• Digi al ans o ma ion
• Capaci y building o Road Sa e y p o essionals
• Implemen ing hie a chical, sa e and connec ed oad schemes is essen ial in oad sa e y ools
and measu es, ensu ing sus ainabili y is a key elemen in mode nising a el
• Simula ion o a ic dis ibu ion in he e en o empo a y a ic o ganisa ion o he
implemen a ion o se e al oad in es men s a he same ime; managemen o he so-called
g een wa e in a ic ligh s o public anspo
Figu e 4-2 Dis ibu ion o scena io in es iga ion impo ance
phoebe-p ojec .eu
Copy igh © by PHOEBE
112
Deli e able 1.1 – S a e o he A and end-use needs e iew
The nex sessions o he su ey aimed o deep di e in o he PHOEBE hema ic a eas. F om he
Beha iou al models, i seems ha he esponden s end o ag ee ha i is highly impo an o include
human beha iou in mic oscopic a ic simula ion (38%), and in a oad assessmen ool (36%), as well as
de i ing human beha iou pa e ns om elema ics da a (23%) a e highly impo an . Responden s also
indica ed he ypes o beha iou s ha PHOEBE should explo e o d i e s, pedes ians and cyclis s (Table
4-1)
Use Type
Sugges ed beha iou s
D i e s
Headways, gaze, ehicle posi ion, accele a ion, ac ual speed on junc ions, eac ion
imes, dispe sion o speeds wi hin he a ic low
Le el o imp udence, speeding, dange ous manoeu es, agg essi e d i ing, dis ega d
o a ic ules, d i e a igue, agg essi e d i ing, la e lane change, no s opping on
pedes ian c ossing, ed
ligh unning, d i ing on public anspo lane, iola ions o
signage o yielding o pedes ians/bikes
Pedes ian (and VRU) espec , espec o speed limi s and signs
All kind o dis ac ions e.g. d ink d i ing, mobile phone dis ac ion
Poo isibili y issues due snow, ice, la ligh condi ions (ze o con as s condi ion)
D i e esponsibili y on he oad o himsel and o he oad use s
Connec ions be ween d i ing (mic o-)pa e ns
Look a di e en mood and hus beha iou : espec i e; p ac ically espec i e;
dange ously espec i e, d i ing beha iou pa e n: speed/ a ic code
compliance/d inking habi s
Legal/illegal pa king, pa king on igh side o lane o 5-15 minu es (quick shopping,
bake y, lo e y e c.), walking dis ance om pa king o des ina ion, inding pa king a ea
Whiche e ones ha e he g ea es impac on a ic low/in e ac ions
How o en does a ca d i e check his ca mi o s?
P e en i e beha iou
Pedes ians
Mobile phone dis ac ion (messages, social ne wo ks, e c.), use o headphones on public
oads while walking
Jaywalking (c ossing midblock when no allowed), c ossing wi h e y li le ime le on he
pedes ian cycle, ed ligh iola ions, mid-
block c ossing, walking on bicycle pa h,
c ossing oad nea (10-20m) pedes ian c ossing
(especially when pedes ian lea es
public anspo )
Visibili y (high- is ma e ials), poo isibili y issues due snow, ice, la ligh condi ions (ze o
con as s condi ion)
Hea a e, gaze, posi ion
Le el o imp udence
Look a di e en mood and hus beha iou : espec i e; pa ially espec i e; dange ously
espec i e
Walking dis ance; wai ing ime o c oss
Age, mobili y disabili ies, a en ion o he su oundings, alone/g oups
How a ahead o ime, does a ca d i e look a a pedes ian c ossing an in e sec ion?
Rela ionship wi h he cyclis s and cycle lanes
In eg a ing walking wi h public anspo , bus s ops; dis ance walking
P e en i e beha iou
phoebe-p ojec .eu
Copy igh © by PHOEBE
113
Deli e able 1.1 – S a e o he A and end-use needs e iew
Use Type
Sugges ed beha iou s
Cyclis s
Compliance wi h a ic egula ions: speeding, ed ligh iola ions, c ossing oad whe e e
hey wan , iding on pedes ian pa h o whe e e hey wan o e en when he e is bicycle
pa h, compliance o signage (e.g. s op/yield signs) o yielding o pedes ians, dange ous
manoeu es; agg essi e d i ing; dis ega d o a ic ules; no ci cula ing on cycle pa hs
when he e a e
Dis ac ion: mobile phone dis ac ion (messages, social ne wo ks, e c.), use o
headphones on public oads while cycling, ea phones use and ixa ion,
Speed, headways, accele a ion pedes ian(and o he cyclis s) espec , a igue, helme
use, hea a e, gaze, ehicle posi ion, age
Riding wi h isibili y, poo isibili y issues due snow, ice, la ligh condi ions (ze o
con as s condi ion)
De ian beha iou , ca eless lea ing o mic o scoo e s in he mids o walkways
Look a di e en mood and hus beha iou : espec i e; pa ially espec i e; dange ously
espec i e
Lack o p o ec i e equipmen , sa e y measu es
P e en i e beha iou
Pa king places
Table 4-1 Sugges ed beha iou s PHOEBE should conside o assess.
Mos esponden s also end o ag ee ha new da a collec ion me hods inclusion in a mic oscopic a ic
simula ion and in a oad assessmen ool as well as inclusion o da a ob ained by a elema ics eco ding
de ice a e highly impo an . Speci ically, 27% o he esponden s s a ed ha i is e y impo an in a
mic oscopic simula ion o include new da a collec ion me hods, while 34% and 31% s a ed ha i is e y
impo an in a oad sa e y assessmen ool and in a elema ics eco ding de ice espec i ely. S akeholde s
also highligh ed he ypes o da a ha can be use ul in hese models:
• came as, elema ics, mobile phone da a, loa ing ehicle da a, senso s (e.g. a ligh ing poles),
AI, iRAP
• Mic o mobili y o igin-des ina ion, passenge o igin-des ina ion, pedes ian demand
• Floa ing ca da a (p obe da a), ehicle speed, ehicle ype di ision
• Vehicula con lic s and in e ac ions wi h ulne able use s
• Beha iou al da a ela ed o use s o he oad en i onmen , be hey pedes ians, d i e s and
use s o so modes o mobili y, while using he oad
• Walking ips, bike ips
• Quali y o he da a p o ided (and ou unde s anding o ha da a and how i could in eg a e wi h
ou cu en analysis ools)
Again, mos esponden s end o ag ee ha modal shi e ec s inclusion in a mic oscopic a ic simula ion
(38%) and in a oad assessmen ool (32%) as well as moni o ing modal shi e ec s using a elema ics
eco ding de ice (36%) a e highly impo an . The mos impo an mode shi s o explo e a e:
• F om passenge ca s o Public T anspo
• F om passenge ca s o smoo h modes o mobili y: cycling, walking and sha ed mobili y
se ices
• F om public anspo o cycling
• Mic o pe spec i e
phoebe-p ojec .eu
Copy igh © by PHOEBE
114
Deli e able 1.1 – S a e o he A and end-use needs e iew
• To/ om s a egic / longe dis ance highway ips
• Inc ease o bike sha e
• Incen i ize non-mo o ized modes and ansi
• Sa e Sys em
Some ques ions a e p esen ed bu asking o Socioeconomic models. Mos esponden s end o ag ee
ha socioeconomic pa ame e s inclusion in a mic oscopic a ic simula ion and in a oad assessmen ool,
al hough he pe cen age a e lowe han he p e ious hema ic a eas (22% and 24% espec i ely). He e he
sugges ions a e o explo e:
• Age, gende , educa ion le el, e hnici y, educa ion le el, amily size
• Owne ship o ehicle, dis ance home-wo k, a i udes/pe cep ion, a ea o esidence, anspo
mode
• Local GDP, pu chasing powe , a e age amily income, alue o ime, alue o li e and inju y,
ope a ion cos s
• Nega i e ex e nali ies due o cons uc ion o ca usage
• Space o pedes ians, bikes, bus lanes, and also deli e y
• Concep o 'le elling up' (pe haps some pa ame e s ha could in o m ha )
• Socio-economic inequali y
In he case o he “Road sa e y assessmen ”, when asked abou he impo ance o including
in as uc u e- ela ed sa e y isk in a mic osimula ion ool, he s akeholde s di ided hei opinions in o
sligh ly impo an (39%) and e y impo an (30%), bu he majo i y ag ee ha use elema ics da a is
undamen al o en ich oad sa e y assessmen s (32%). Responden s sugges ed he ollowing elemen s as
essen ial o be conside ed in oad sa e y assessmen s.
• Speed limi s, lane wid h, sidewalk wid h, oad su ace, ho izon al and e ical signs, cu a u e,
isibili y, a ic olume, a ic speed (V85)
• Whe he S ee Ligh ing makes a di e ence
• U ban oad ne wo k isk assessmen
• VRU, ulne abili y-based, pedes ian c ossing
• Walking and Cycling
• Me hodology de ined in he ISO 31000 S anda d
The endency emains o a ic mic osimula ion inclusion in a oad assessmen ool (33% as e y
impo an ) and elema ics da a inclusion in a mic osimula ion ool (27% as e y impo an ). When asked
abou wha should be simula ed, s akeholde s sugges ed:
• In e sec ion mic o-simula ion - pedes ian
• Lane changing in highways
• Di e en ia ing ype o ehicle
• Con lic s
• De ec ion sys ems o speed and a ic densi y
A e anking impo ance o in eg a e he di e en PHOIEBE hema ic a eas, s akeholde s we e asked o
epo hei p o essional needs and he use ulness o ha ing hese in eg a ions. The comple e esul s can
be ound in Annex C. I seems ha mos s akeholde s see ele ance o hei needs o en iching a ic
simula ion wi h in as uc u e sa e y in o ma ion (30%), wi h modal shi in o ma ion (28%), wi h induced
phoebe-p ojec .eu
Copy igh © by PHOEBE
115
Deli e able 1.1 – S a e o he A and end-use needs e iew
demand models (36%) and wi h human beha iou models (26%). In addi ion, mos esponden s s a ed ha
i is e y impo an o imp o e accu acy o a ic simula ion (40%) and o en ich a ic simula ion wi h all
hese a o emen ioned aspec s (34%). Rega ding oad assessmen , mos s akeholde s s a ed ha i is e y
impo an o en ich oad assessmen wi h a ic mic osimula ion in o ma ion (26%) and wi h AI/ML models
(28%). On he o he hand, a sligh impo ance is s a ed om he 24%, 28% and 26% o esponden s in case
o he need o oad assessmen enhanced wi h modal shi in o ma ion, wi h induced demand models and
wi h human beha iou models espec i ely. Finally, 32% o s akeholde s s a ed ha he need o en ich oad
assessmen wi h all hese aspec s is e y impo an .
In he case o use ulness, mos s akeholde s end o ag ee ha i would be e y use ul in e e yday asks o
ha e a PHOBE-de eloped ool ha could en ich a ic simula ion wi h in as uc u e sa e y in o ma ion
(24%) and wi h modal shi in o ma ion (22%). On he o he hand, a mode a e use ulness is s a ed om he
22% and 24% o esponden s in case o a PHOBE-de eloped ool ha could en ich a ic simula ion wi h
induced demand models and wi h human beha iou models espec i ely. Rega ding oad assessmen , he
majo i y o s akeholde s s a ed ha a PHOEBE-de eloped ool would be sligh ly use ul i could en ich oad
assessmen wi h mic osimula ion in o ma ion (22%), wi h modal shi in o ma ion (26%), induced demand
models (26%) and wi h human beha iou models (22%). Finally, a ool ha could enhance oad assessmen
wi h AI/ML models and wi h all he a o emen ioned aspec s would be e y use ul o he majo i y o
s akeholde s (28% and 26% espec i ely).
The wo inal ques ions aimed o unde s and wi h wha equency a use will use he PHOEBE F amewo k
and i hey belie ed he amewo k could impac he ac ual numbe o c ash da a. A ound 70% o he
esponden s ag ee ha hey will use he expec ed PHOEBE-de eloped ool om o en enough o e y
equen ly. S akeholde s a e mo e scep ical abou he eal impac o he ool, wi h 58% o he esponden
eplying wi h a es a ying om will impac enough o an ex eme impac .
4.1.1 Use case s akeholde s examina ion
In o de o highligh any possible di e si y o s akeholde s opinion o he h ee di e en PHOEBE use cases
(A hens, Valencia and Wes Midlands), simila analyses we e gene a ed o each use case g oup o
s akeholde s. Fo his eason, he answe s o he s akeholde s wo king in A hens, Valencia and Wes
Midlands we e g ouped esul ing in se en s akeholde s in he g oup o A hens, i e in Valencia and h ee
in Wes Midlands egion. I should be acknowledged ha he pa icula in es iga ions p esen ed in his
chap e a e an e o o glean speci ic needs and indi idual cha ac e is ics o each use case ci y a he han
gene al-use ans e able s a is ics. Fi s o all, he e a e no di e ences ega ding educa ion le els, since
he Speci ically, he majo i y o s akeholde s hold a mas e deg ee o all he h ee g oups (71% o A hens,
60% o Valencia and 67% o Wes Midlands s akeholde s). On he o he hand, he ools hey use in hei
daily ac i i ies seems o a y among he g oups as showed in Figu e 4-3.The majo i y (17%) o A hens
s akeholde s use a ic signal op imiza ion so wa e, while he majo i y (21%) o Valencia s akeholde s use
ou ing so wa e and he majo i y (25%) o Wes Midlands s akeholde s use mac oscopic a ic simula ion.
Figu e 4-3 Dis ibu ion o so wa e/ ool ype usage pe use case s akeholde s
phoebe-p ojec .eu
Copy igh © by PHOEBE
116
Deli e able 1.1 – S a e o he A and end-use needs e iew
In Figu e 4-4, he impo ance o di e en scena ios in es iga ion is p esen ed. Acco ding o hese indings,
i seems ha he h ee di e en g oups o s akeholde s end o ag ee ha i is highly impo an o in es iga e
he majo i y o he scena ios. On he o he hand, he e a e some cases whe e i can be no iced ha he
opinions o he h ee di e en g oups o s akeholde s a e di e en ia ed. Speci ically, A hens s akeholde s
s a ed a mode a e impo ance in case o in es iga ing he implemen a ion o egula o y measu es o limi
speeds (29%), he managemen o con lic due o in oduc ion o new anspo modes (29%), he
implemen a ion o newe s anda ds o bike lanes design (29%) and mode nising sus ainable a el aligning
wi h exis ing and p edic ed demand in ad ance o majo e en s (43%), while he es g oups epo ed a
highe impo ance. Fu he mo e, Wes Midlands s akeholde s s a ed a mode a e impo ance o
in es iga ing p omo ion o public anspo (33%), while A hens (57%) and Valencia (80%) s akeholde s
s a ed ha his kind o in es iga ion is e y impo an .
The impo ance o in eg a ing he hema ic a eas also a ies be ween use cases. Fo A hens, s akeholde s
in eg a ing modal shi and socioeconomic pa ame e s in o mic oscopic simula ion ools, socioeconomic
pa ame e s in o oad sa e y assessmen and human beha iou in elema ic eco ding de ices a e he mos
a ed in eg a ions. In he case o Valencia, he highes pe cen age o esponses a e dis ibu ed among
human beha iou and socioeconomic pa ame e s in o mic oscopic simula ion ools and modal shi in o
oad sa e y assessmen ools. The educed numbe o esponses om he Wes Midlands egions and he
high alue gi en o he in eg a ions sugges ed making i challenging o de e mine he mos alued
in eg a ion. Ne e heless, o he gi en esponses, in eg a ing modal shi is impo an o all he
esponden s. The same applies o human beha iou in eg a ion wi h oad sa e y assessmen and elema ic
eco ding de ices.
Table 4-2 In eg a ion in o PHOEBE hema ic a eas conside ed e y impo an o he use case s akeholde s
Fu he mo e, no signi ican di e en ia ion be ween he opinions o he di e en s akeholde g oups is
no iced o he es o he aspec s e alua ed. Speci ically, mos o all s akeholde g oups s a ed a high
impo ance o he ollowing cases: in as uc u e sa e y isk assessmen me hodologies in a mic osimula ion
ool, elema ics da a in in as uc u e sa e y isk assessmen me hodologies, mic osimula ion ools in a oad
assessmen ool and elema ics da a in a mic osimula ion ool.
phoebe-p ojec .eu
Copy igh © by PHOEBE
117
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 4-4 Dis ibu ion o scena io in es iga ion impo ance pe use case s akeholde s
phoebe-p ojec .eu
Copy igh © by PHOEBE
118
Deli e able 1.1 – S a e o he A and end-use needs e iew
Acco ding o he o e all p o essional needs, i seems ha he h ee di e en g oups o s akeholde s end
o ag ee on he needs ha a e highly impo an . Ne e heless, he e a e ew cases whe e i can be no iced
ha he opinions o he h ee di e en g oups o s akeholde s a e di e en ia ed. Speci ically, Valencia
s akeholde s s a ed a mode a e impo ance o he need o enhancing oad assessmen wi h a ic
mic osimula ion, wi h modal shi in o ma ion, wi h induced demand models, wi h human beha iou models
and wi h AI/ML models as well.
In he case o he use ulness o a PHOEBE-de eloped ool he majo i y o Wes Midlands s akeholde s end
o ag ee ha all kinds o he possible ea u es o a PHOBE-de eloped ool would be e y use ul in hei
e e yday asks. On he o he hand, a mode a e use ulness is s a ed om he Valencia s akeholde s o all
he p oposed capabili ies o a ool. Focusing on he answe s gi en om he A hens s akeholde s, hey
ag eed wi h Valencia s akeholde s o ea u es o en iching a ic simula ion wi h induced demand models,
oad assessmen wi h modal shi in o ma ion, wi h induced demand models and wi h AI/ML models as well.
Fo he es o capabili ies, A hens s akeholde s s a ed a highe use ulness ag eeing wi h Wes Midlands
s akeholde opinion.
Mos esponden s (43% om A hens, 40% om Valencia and 30% om Wes Midlands) s a ed ha hey
will use he expec ed PHOBE-de eloped ool o en enough. Rega ding he e ec i eness o he expec ed
PHOEBE ool on eal c ash numbe s, i seems ha he majo i y (67%) o Wes Midlands s akeholde s a e
qui e op imis ic s a ing an adequa e impac , while he majo i y (43%) o A hens ones poin ed o a mo e
mode a e impac . In he case o Valencia s akeholde s, a mo e pessimis ic poin o iew is no iced, as he
40% s a ed ha i will no impac a all he eal c ash numbe s.
The comple e esul s o he use case s akeholde s a e p esen ed in Annex D: Addi ional s akeholde s
su ey esul s g aphs In closing, some addi ional commen s ha e been p o ided by he in e iewe s
ega ding he in e iews wi h Valencia Ci y Hall, and speci ically:
• In e ac ions be ween cyclis s and pedes ians a e a cons an conce n o he Valencia Ci y Hall.
• The Valencia Ci y Hall conside s ca s as he s onges dange o bo h pedes ians and cyclis s,
hus hey a e commi ed o educing hei p esence and can be in e es ed in ools ha help
hem p o e his ac .
• Road sa e y is conside ed e y impo an and an u gen ma e by ETRA EMT and ATMV,
especially when ela ed o he in e ac ions be ween cyclis s and pedes ians.
• The e is a g ea in e es (Valencia Ci y Hall, EMT, ATMV) in boos ing modal shi ) connec ion
public anspo and non-mo o ised ehicles.
• A g ea in e es has been shown in in oduc ion o AI in da a collec ion me hods.
• Fo wha ela es o socio economic analysis: I s inclusion in mic oscopic analysis has no been
conside ed use ul by he majo i y o he in e iewees, bu hey would include his kind o
analysis when connec ing ype o anspo mode o a ea o esidence, and e enues pe yea ,
o be e unde s and co ela ions.
Acco ding o EMT, i would be in e es ing o s udy di es men in public anspo and i s edi ec ion (i any)
owa d non-mo o ised public anspo a ion op ions.
4.2 Focus g oups
F om he i e sessions o he ocus g oups, key issues we e iden i ied. These issues a e linked and o e lap
wi h one ano he , and hey appea ed h oughou all he ques ions, no ollowing a linea discussion, as
expec ed.
Key issues iden i ied:
• Road Sa e y Unde s anding
phoebe-p ojec .eu
Copy igh © by PHOEBE
119
Deli e able 1.1 – S a e o he A and end-use needs e iew
• O ganisa ional s uc u es (knowledge and p ac ices)
• O ganisa ional esou ces and capabili ies
• Concep s eali y
The Figu e 4-5 demons a es how hese issues a e connec ed and in luence one ano he .
Figu e 4-5 Key issues iden i ied in ocus g oups and hei in e dependencies
4.2.1 Road Sa e y Unde s anding
Among he opics ha eme ged om he ocus g oups, one essen ial and unde lying issue is Road sa e y
unde s anding, which in luences many cu en p ac ices and challenges aced by p ac i ione s.
P o essionals alked abou he di e en ways and pe spec i es o oad sa e y, wi h “issues on how we
app oach oad sa e y”.
Pa icipan s see a de iciency in how he link be ween oad dea hs and o he decisions ( owa ds o he
modes) made in anspo and mobili y a e pe cei ed, ela ing o poli icians’ eal unde s anding o ision
ze o. When he inancial and poli ical easons o oad sa e y weigh in decisions (pe cei ed e sus eal
isk), i can lead o misleading o dange ous p ocesses since poli icians do no ha e a p ope unde s anding
o oad sa e y. When is a inancial decision, he p io i y will be on he e u n on in es men . When i is a
poli ical decision, he impo ance is o please someone o a g oup.
Tha issue also ela es o he o ganisa ional s uc u es ha will be discussed below, which lead ci ies, and
ci izens no o ha e he ull poin o iew. I was a gued ha showing he wide bene i s o oad sa e y is
needed and ha a ool ha could suppo p o essionals in doing so would be use ul.
How o show ha a lack o oad sa e y indi ec ly impac s o he ci y goals like sus ainable anspo , li eliness
and a ac i eness o he ci ies? P o essionals ag eed ha compa ing impac s on collision and oad sa e y
phoebe-p ojec .eu
Copy igh © by PHOEBE
120
Deli e able 1.1 – S a e o he A and end-use needs e iew
is cu en ly p oblema ic. i.e. o compa e he numbe o sa ed li es and he inc ease in a el ime,
“conges ion is angible. Can we calcula e a ic sa e y?”.
“i would be easie o say we will sa e i e li es
a he han i e minu es om you commu e”.
P o essionals aised he issue ha he cos -bene i analysis o oad sa e y measu es should b ing isibili y
o he social cos s and he economic and heal h bene i s. A mo e comp ehensi e cos -bene i analysis could
minimise bad pe cep ions ha could a ise, mainly linked o educing speeds, a ic lows, ealloca ion o
space and o he measu es. The e should be possibili ies o enable poli icians o show ha he implemen ed
solu ions a e he bes way o go.
E en hough i is ele an ha oad sa e y needs o be economically iable, comp ehensi e oad sa e y
in e en ions do no necessa ily mean a conside able cos inc ease. Howe e , conce ning he poli ical
cos s, o example, conges ion, o o he unpopula decisions, whe e poli icians ea he complain s,
pa icipan s a gued ha oad sa e y needs o be made mo e isible o p e ail. The popula ion is no ully
awa e o oad sa e y and wha i migh en ail. Thus, he e also needs o be an ine i able shi in socie y in
ela ion o oad sa e y. Pa icipan s no ed ha people a e o en “ alking abou pe cep ions, bias
pe cep ions”.
“Acciden s wi h ca s a e pe cei ed as no mal.”
The ade-o s make he con e sa ion and decision-making p ocesses di icul , o example, when i in ol es
emo ing pa king, changing a ic lows, o implemen ing one-way s ee s. Mo eo e , because ci ies and
ci izens a e no ha ing a comple e pic u e o all ha implies in aking measu es o inc eased oad sa e y,
especially o VRUs.
“No one is agains oad sa e y, bu he ade-o s make he decision complica ed.”
So, wha could help wi h hese ade-o s? I was highligh ed ha igu ing ou he p oblem i s and
es ablishing p io i ies is impo an . So, wha is he eal p oblem, and wha is he ci y ying o ackle?
Pa icipan s alked abou a lack o me hodology o p io i isa ion – lack o e alua ion me hods o di e en
use s – p io i isa ion and isk e alua ion can s ill be linked wi h di e en lobbying (ca s, ucks, buses, bikes).
They also discussed ha cu en ools and p ocesses o oad sa e y do no accommoda e he coexis ence
among di e en use s. One pa icipan no ed ha “you hink you a e doing i sa e o pedes ians, bu i is
no ”, because p ocesses a e s ill e y much ca - ocused and using ca -o ien ed solu ions o VRUs.
“We wo k wi h he in en ion o inc easing oad sa e y bu some imes goes
by he o he way.”
The issues on how oad sa e y is unde s ood and pe cei ed aise ques ions such as is oad sa e y cu en ly
biased? A e VRUs being ully conside ed in oad sa e y decisions? The pa icipan s o he ocus g oups
p o ided mo e insigh s in o hese ques ions ha will be discussed u he . Howe e , i is clea ha hese
issues a e mo e deeply oo ed and ha hey o e lap and e lec in he o he key opics iden i ied.
4.2.2 O ganisa ional s uc u es (knowledge and p ac ices)
Looking in o he O ganisa ional s uc u es (knowledge and p ac ices), i is possible o ecognise ha
hese s uc u es a e no disassocia ed wi h he o e all unde s anding o oad sa e y and he e y own
capabili ies o o ganisa ions, and hese will also g ea ly in luence he cu en oad sa e y p ocesses and
decisions and migh pose challenges in he accep ance o u u e app oaches. Pa icipan s also conside ed
he de iciencies in he cu en ools and he assessmen o VRUs and new mobili ies wi hin he
o ganisa ional knowledge and p ac ices.
phoebe-p ojec .eu
Copy igh © by PHOEBE
127
Deli e able 1.1 – S a e o he A and end-use needs e iew
4.3.3 Me hodological and echnical needs s a emen s
• In o de o induce modal shi , he e is a need o backcas , looking a he a ic and sa e y
e ec s gi en he selec ed policy goal e.g. a goal o ealloca e he space o cyclis s.
• I is impo an o acknowledge ha VRU beha iou is mo e andom, i canno be modelled on
he o igin-des ina ion ajec o y way like ca s.
• The andom and unp edic able way o mo emen o pedes ians should be seen as a
means o ge an e icien pa h gi en he de iciencies o he acili ies, and no necessa ily as a
eckless and iola ing beha iou .
• The ollowing human ac o s a e p io i ized by he consul ed s akeholde s:
o Dis ac ion (all ypes, incl. mobiles and social ne wo ks)
o Speeding, compliance
o Jaywalking, pa h iola ion
o Agg essi eness
• The ollowing new da a sou ce a e p io i ized by he consul ed s akeholde s:
o Sma phones, came as, senso s’ da a
o Floa ing ca da a
o O igin-des ina ion da a o mic omobili y
o Telema ics da a ha can be usable o isk assessmen :]
• A he same ime, new da a sou ces migh o e whelm local au ho i ies, in he lack o
human esou ces ha would be ained and knowledgeable on using hem
• Da a accessibili y and he high cos o wo k wi h he eme ging da a is a ba ie o hei
exploi a ion by ci ies. I is no clea o many pa icipan how da a on VRU mo emen s can be
ob ained o e en sys ema ically collec ed.
• Pa icipan s exp essed he need o opening a dialogue wi h ope a o s o mic omobili ies in
o de sha e da a, and indica ed ha i would be an added alue o PHOEBE o demons a e
he use o such da a.
4.3.4 PHOEBE use cases
While he e was b oad consensus among he su ey pa icipan s, a numbe o poin s can be highligh ed in
e ms o he pa icula i ies o he h ee dis inc PHOEBE use case ci ies.
• A hens s akeholde s s a ed a mode a e impo ance in case o in es iga ing he implemen a ion
o egula o y measu es o limi speeds, he managemen o con lic due o in oduc ion o new
anspo modes, he implemen a ion o newe s anda ds o bike lanes design and mode nising
sus ainable a el aligning wi h exis ing and p edic ed demand in ad ance o majo e en s,
while he es g oups epo ed a highe impo ance.
• Wes Midlands s akeholde s s a ed a mode a e impo ance o in es iga ing p omo ion o
public anspo , while A hens and Valencia s akeholde s s a ed ha his kind o in es iga ion
is e y impo an .
• A hens and Valencia s akeholde s end o ag ee ha is mode a ely impo an o include human
beha iou in a oad assessmen ool as well as o de i e human beha iou pa e ns om
elema ics da a, while Wes Midlands s akeholde s opinion is ha o bo h cases is e y
impo an .
• A hens and Wes Midlands s akeholde s end o ag ee ha i is highly impo an o include
new da a collec ion me hods in a mic osimula ion and oad assessmen ool as well as da a
phoebe-p ojec .eu
Copy igh © by PHOEBE
128
Deli e able 1.1 – S a e o he A and end-use needs e iew
ob ained by a elema ics eco ding de ice. On he o he hand, he Valencia s akeholde s s a ed
a mode a e impo ance o all hese cases.
• O e all Valencia s akeholde s appea ed o ha e mo e mode a e iews o he o e all
p o essional needs when i comes o mos o he aspec s o be in eg a ed in he PHOEBE
amewo k.
These pa icula i ies do no e lec majo disag eemen s, hey do highligh he need o a ailo ed app oach
in each use case. O e all, i is impo an o make su e ha es ed use cases a e aligned wi h ac ual
needs and p io i ies – also in ligh o da a a ailabili y. This can be seen as a challenge and an
oppo uni y o de elop a holis ic bu case-sensi i e ool. Alongside PHOEBE’s pilo s, he amewo k
should con inue an hones and open discussion wi h ci ies. This will ensu e ha he p ojec ou comes a e
applicable and ele an o eal-wo ld scena io, making he in e en ions and amewo k mo e e ec i e and
impac ul.
Undoub edly wha he PHOEBE F amewo k can achie e has a limi a ion since such a shi encompasses
many o he aspec s beyond he p ojec scope. The ocus g oups’ pa icipan s aised many ques ions and
challenges ha PHOEBE ce ainly will encoun e . The eali y o ci ies and public au ho i ies is illed wi h
esis ance and lack many esou ces, making i challenging o hem o adop new ways o wo king e en
when he e is he in en ion o. Thus, o a oid alse expec a ions and ensu e mo e e ec i e ou comes, i is
impo an ha PHOEBE ully ecognises i s limi a ions, no as sho comings, bu as le e age as a key playe
in he oad sa e y discussion, ad oca ing o he necessa y changes and new ways o add ess hem.
Alongside PHOEBE’s pilo s, con inuing wi h an hones and open discussion wi h ci ies and o he
s akeholde s is essen ial o ha . The ocus g oups we e an ini ial s ep in his p ocess o acknowledging
hei needs and will con inue h oughou he p ojec in many ways.
phoebe-p ojec .eu
Copy igh © by PHOEBE
129
Deli e able 1.1 – S a e o he A and end-use needs e iew
5 Conclusions and ecommenda ions
The PHOEBE F amewo k is a me hodological app oach designed o ci ies o imp o e unde s anding o
he sa e y implica ions o u u e changes in he anspo sys ems, such as beha iou al changes, edesign
o new in as uc u e, o he in oduc ion o a new mode. To do so, i aims o b ing oge he se e al exis ing
elemen s including:
• T a ic simula ion en i onmen s, which p o ide insigh s in o how hese a ious elemen s
in e ac in i ual a ic scena ios
• Road sa e y assessmen , which can p o ide objec i e measu es o oad use sa e y linked o
speed, low and he physical oad ea u es
• Human beha iou models which can help an icipa e how oad use s espond in gi en
scena ios, how people selec which mode o anspo o use o when/whe e o ake a ip ( he
la e coming in o he ca ego y o mode shi and induced demand models, which can p edic
consume choices and oad use low changes), and
• Socio-economic impac assessmen which can unde s and he heal h, sa e y, economic and
en i onmen al impac s o changes in he anspo sys em.
Toge he , hese elemen s will enable analysis on he combined e ec o u u e changes, as well as how
his will change o e ime.
This deli e able, which is he di ec ou pu o Task 1.1 o he p ojec , aims o e iew he s a e-o - he-a in
each componen u ilized in he PHOEBE F amewo k, and gain a deepe unde s anding o he needs o
ci ies and anspo planne s. The indings o he s a e-o - he-a e iew will be o in o m he heo e ical
de elopmen , including he key p inciples and me hodology o he PHOEBE F amewo k (Task 1.2). The
Use Needs S a emen will in o m he de elopmen and design o ools and suppo ma e ials o end use s
o ensu e he amewo k can be u ilised o i s s a ed pu pose.
The de elopmen o he PHOEBE F amewo k in ol es h ee p incipal aspec s:
1. Fi s is he heo e ical de elopmen , which includes he esea ch, whe e necessa y, o key p inciples
and me hodology o he amewo k
2. Second is he echnical de elopmen s equi ed o ensu e ha he echnical componen s o he
amewo k ha e he necessa y unc ionali y o mee he amewo k’s needs and p oduced he analysis
equi ed, and
3. Thi d a e he ools and suppo ma e ials o end use s o ensu e he amewo k can be u ilised o
i s s a ed pu pose.
The ollowing sec ions b eak down he discussion and implica ions o he e iew and s akeholde
consul a ion indings o each aspec .
5.1 Implica ions on he heo e ical de elopmen o he PHOEBE
F amewo k
The de elopmen o he PHOEBE me hodology ela es di ec ly o he i s objec i e o he PHOEBE p ojec ,
which is o de elop a new, eplicable me hodology o dynamic sa e y p edic ion and socio-
economic e alua ion. This me hodology needs o:
• Be bo h ha monised and scalable o all ypes o oad use s
• Link app op ia e isk indica o s a mic o-, meso- and mac o-le el
phoebe-p ojec .eu
Copy igh © by PHOEBE
130
Deli e able 1.1 – S a e o he A and end-use needs e iew
• Con ain eplicable me hods o modelling and simula ing modal shi and isk o u u e new
modes, beha iou al adap a ion and oad use sa e y, and
• Allow he es ima ion o social cos s and bene i s in e ms o heal h and sa e y, en i onmen al,
and economic aspec s.
The concep o he PHOEBE F amewo k is o es ablish a me hodological amewo k a ound he in eg a ed
applica ions o a ic simula ion ools, oad sa e y assessmen and da a.
The e iew highligh ed he pi o al ole o human beha iou models in a ic simula ion ools, pa icula ly o
he inclusion o ulne able oad use g oups, such as pedes ians and bicyclis s, which elies on he
simula ion o mo e ealis ic oad use beha iou s. Human beha iou is also unde s anding demand,
including mode choice and mode shi .
Human beha iou is cu en ly cap u ed in oad sa e y assessmen models, such as he S a Ra ings,
p ima ily h ough ope a ing speeds and oad use low da a, such as he equency o pedes ians c ossing
a a gi en loca ion. Beha iou is also an inhe en pa o he in as uc u e isk ac o s, which a e based on
es ablished c ash modi ica ion ac o s (CMF). CMF cap u e oad use beha iou s change in esponse o
in as uc u e. Fo example, whe e d i e s slow down in esponse o na owe oad lanes.
This e iew con i med ha he S a Ra ings a e an app op ia e oad sa e y assessmen me hodology o
mee he needs o he p ojec aims. This me hod inco po a es indi idual models o VRUs, and is applicable
ac oss all egions and oad ypes (including u ban oad ne wo ks). Howe e , S a Ra ings p o ide s a ic
sa e y a ings. To allow o dynamic sa e y p edic ion, and a be e ep esen a ion o human beha iou ,
speed and low a iables om a ic simula ion models will be inco po a ed.
Fo he socio-economic e alua ion, he e iew iden i ied se e al me hodologies, all wi h hei ad an ages
and limi a ions. In Task 1.3, a mo e in-dep h e iew in o p e ious wo k will be ca ied ou in e ms o he
heo e ical amewo k o sa e y impac s on he socio-economics o he di e en VRUs and he di e en
popula ion g oups. Di e en di ec and indi ec impac s o he sa e y ex e nali ies will be iden i ied and he
measu es o quan i y hem. I is no ed ha his e alua ion is ini ially conside ed as a pos -hoc ask o he
sa e y and a ic ou pu s o he enhanced simula ion models.
E e y componen o he PHOEBE amewo k elies on no el da a applica ions, and u ilisa ion o hese da a
sou ces is a key pa o he me hodological app oach. The e a e a numbe o success ul expe iences wi h
using hese da a sou ces, enabling se e al impo an messages o be aken as ega ds he PHOEBE
F amewo k:
• A i icial Neu al Ne wo ks a e e y app op ia e o sho e m a ic o ecas ing, whe eas deep
lea ning models can be success ul o longe e m o ecas ing
• Fo ecas ing he change in oad isk would equi e Bayesian Belie Ne wo ks (BBN) o o he
ad anced p obabilis ic models
• Open pla o ms like OSM should be used o ne wo k iden i ica ion, despi e he he e ogenei y
o accu acy – gi en ha hei use is no cos ly, hey should be p omo ed as a i s line da a
sou ce.
I is impo an o dis inguish he da a needs o he di e en PHOEBE componen s o he me hodology.
The PHOEBE p ojec aims o ad ance he s a e-o - he-a o each o i s componen s in some way. Bu he
ue ad ance o he s a e-o - he-a is he in eg a ion o a ic simula ion and oad sa e y assessmen which
ha e adi ionally been used in isola ion o each o he , and o di e en pu poses. Achie ing a
me hodological amewo k which can u ilise hese componen s and hen using his o model mode shi and
induced demand, and calcula e socio-economic impac s equi es linking he mic o- (indi idual ac ions),
meso- (g oups o oad use s) and mac o-le els ( he ne wo k). How hese le els connec will need o be
ca e ully conside ed in Task 1.2 and desc ibed in Deli e able 1.2, he Theo e ical p inciples and
me hodological app oach o he PHOEBE amewo k and selec ion o ools (due in mon h 12).
phoebe-p ojec .eu
Copy igh © by PHOEBE
131
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 5-1 Depic ion o he mul i-le el analysis p oposed by he PHOEBE amewo k
5.2 Implica ions on he echnical de elopmen o he PHOEBE
F amewo k
Se e al echnical componen s will be de eloped as pa o he PHOEBE p ojec o suppo he PHOEBE
F amewo k and demons a e and acili a e i s use.
The de elopmen o echnological aspec s ela es di ec ly o objec i es 2, 3 and 4 o he PHOEBE p ojec ,
which a e:
• To de elop a oad sa e y module in he a ic simula ion so wa e which employs ha monised
sa e y de ini ions
• To de elop enhanced and in eg a ed u ban isk assessmen models and ools o he
applica ion o he me hodological amewo k h ough op imised model design, enhanced isk
ac o s and dynamic inpu s and ou pu s o u ban sa e y assessmen s o challenging scena ios
in u u e oad a ic, and
• To embody social componen s in o isk assessmen s so ha changes in human beha iou , and
mode and ip choices can be be e unde s ood o hei consequences on sa e y.
The PHOEBE amewo k includes a numbe o in e sec ing componen s ha need o be enhanced,
upg aded and in e linked o achie e he scien i ic objec i es o he me hodological amewo k and esul in
a usable and e icien policy suppo ool. The esul s o he p esen s a e-o - he a highligh ed he exis ing
scien i ic and policy gaps. Mo eo e , hey shed ligh on he s ong in e - ela ionship be ween a ic
simula ion, oad sa e y assessmen and he ole o human beha iou models o enhance he c edibili y o
bo h; his ‘ iangle’ lies a he co e o PHOEBE me hodology.
A he same ime, he mode choice and modal shi models can be seen as he i s s ep o be aken in o de
o simula e a ic and sa e y impac s, while socioeconomic impac s a e he inal ou comes o be es ima ed
om he a ic and sa e y ou pu s o he amewo k. The new eme ging da a sou ces a e a ho izon al
dimension ha will enhance he p edic i e and explana o y powe o all componen s o he amewo k.
phoebe-p ojec .eu
Copy igh © by PHOEBE
132
Deli e able 1.1 – S a e o he A and end-use needs e iew
Figu e 6.1 below a emp s o demons a e hese in e - ela ions, as hey a e cla i ied om he in-dep h e iew
o he SoA.
Figu e 6.5-2 Concep ual design o PHOEBE me hodological mewo k
3.5.3 Speci ic issues o be conside ed in he de elopmen o he PHOEBE
echnical aspec s
The e iew iden i ied a ange o issues which would need o be closely conside ed du ing he echnical
de elopmen s o simula ion models (pa icula ly wi h ega d o inco po a ing human beha iou models),
oad sa e y assessmen and mode shi /induced demand modelling.
The e iew o he s a e o he a e ealed ha , while he e is a s ong ela ionship be ween su oga e sa e y
measu es, ypically used in a ic simula ion, and eal c ash da a, ypically a ailable om c ash eco ds,
he e is lack o comple e and holis ic es ablished mechanisms o simula ing po en ial c ashes.
iRAP isk assessmen me hodology and he esul ing s a a ings a e a e y well es ablished and globally
applicable app oach, howe e he es ima es a e s a ic. The isk assessmen me hods need o be u he
de eloped o accoun o he a iabili y o speeds and low. The new eme ging da a sou ces can play a key
ole in ha .
Su oga e measu es o sa e y a e he s anda d way o assessing sa e y ia a ic simula ion. The FHWA
Su oga e Sa e y Assessmen Model (SSAM) mainly uses TTC and PET, and he e a e some s udies wi h
enhanced, angle-based con lic es ima es. Howe e he me hod does no ha e de ailed ajec o ies, and
has been only applied o ehicle- ehicle con lic s. The ne wo k sa e y es ima ion is based on he
agg ega ion o he o al numbe s o con lic s.
Comme cial a ic simula ion ools, e.g. Aimsun, Vissim and SUMO, a e designed in a way ha ehicles
always espec dis ances o o he ehicles and canno e e c ash. This implies ha a ailable a ic
simula ion ools a e designed o be ‘sa e by na u e’. Howe e , s udies ha e shown ha i is possible o
Ca d i e s
Pedes ians
Cyclis s
Mic o mobili y
Mode choice and modal shi modelling
Socioeconomic impac assessmen
S a e o he a and exis ing knowledge / models
Eme ging da a sou ces
Road sa e y assessmen
T a ic simula ion
Beha iou al models
phoebe-p ojec .eu
Copy igh © by PHOEBE
133
Deli e able 1.1 – S a e o he A and end-use needs e iew
simula e c ashes ia hese ools (e.g. by elaxing ce ain assump ions abou he compliance and
‘obedience’ o ehicles and obse ing c ashes as o e lapping ehicle posi ions).
The e a e wo ypes o models: he ‘ope a ional le el’ models, namely ca ollowing (CF) looking a
longi udinal beha iou s, and he ‘ ac ical’ le el models, namely lane changing (LC) looking a la e al
beha iou s. Bo h exis ing models p esen ed so a assume a pe ec d i e wi h pe ec in o ma ion. This
means ha hey do no accoun o any d i e dis ac ion, excessi e isk- aking, ad e se wea he ,
pe cep ion mis akes o biases. The e o e, he e a e conce ns whe he hey indeed ep esen how d i e s
beha e in he eal wo ld. Fo example, empi ical s udies ha e shown much smalle headways and gaps
accep ed han he ypical h esholds used in simula ion ools.
In e ms o simula ion heo ies and models, pedes ian simula ion is mo ing om social o ce models o
mul i-agen models o mo e lexibili y and less compu a ional cos . Again, comme cial packages like
Aimsun and Vissim include agen -based pedes ian plug-ins, bu assume an ideal beha iou o ca s and
pedes ians.
F om he a ic simula ion modelling pe spec i e, in heo y i is possible o in oduce human beha iou ,
he e ogenei y & choices in he d i ing / walking models, bu he p oblem is hei calib a ion and alida ion
– i is expec ed ha in mos cases his is highly challenging due o lack o da a.
The lack o da a on VRUs beha iou is a key ba ie o he in oduc ion o human beha iou in simula ion,
and i s alida ion. Resea ch shows ha human beha iou in e ms o pe cep ion e o s, isibili y e c. is
easie o accoun o in simula ion, han iola ions and isk aking. Human ac o s such as dis ac ion and
a igue show la ge dis ibu ions, and he e a e co ela ions o hese beha iou s wi h speed, headways e c.
The e a e a ew ecen p omising examples o including human ac o s a he ope a ional le el o a ic
simula ion (CF) (Cal e e al., 2020), by means o using a ‘si ua ion awa eness’ and ‘ ask sa u a ion’
amewo k, and i s impac s on e.g. headways. Much o his wo k has been done wi hin he SAFE-UP
Ho izon p ojec . I could be an oppo uni y o PHOEBE o liaise wi h his p ojec and exchange abou hese
de elopmen s.
The e is a di e en unde s anding o he meaning o ‘human beha iou ’. As men ioned, in a ic simula ion
d i ing beha iou models a e mos ly kinema ics-based, and human a ia ions a e also aken om a
‘pa ame e ’ pe spec i e – o ins ance, pedes ian simula ion models ha allow di e en heigh s o
pedes ians. Howe e , om he iewpoin o he discipline o human ac o s in oad sa e y, human beha iou
has a much b oade and mo e complex analysis.
I is well known ha he e is a wo-way ela ionship be ween human ac o s and a ic beha iou ; one aspec
is beha iou al adap a ion, in which d i e s adjus o a ce ain oad sa e y measu e he eby no allowing i s
ull expec ed e ec o be obse ed. Ano he example is endogenei y, o example a igued d i e s
main aining longe headways because hey a e awa e o hei impai men , bu a igue may also be mo e
likely a nigh ime when he e is less a ic and headways a e longe . This “chicken and egg” p oblem
makes i di icul o iden i y which is he dependen a iable o he analysis, and speci ic modelling
echniques a e needed o accoun o his.
The wo-way ela ionship be ween human ac o s and esul ing a ic beha iou needs o be c edibly
ep esen ed in he dynamics o a ic simula ion; his can be a e y complex esea ch ask. The e a e
se e al heo ies and models o human beha iou in a ic. In many cases, human ac o s a e unobse ed
( ela ed o pe cep ions, mo i a ions, pe sonali y ai s) and he e o e a e modelled as la en a iables. In
many cases, su eys a e needed o measu e hese cons uc s; in o he cases, hey can be measu ed by
means o o he obse able indica o s – he e oo new da a sou ces can play a ole, as will be discussed
la e .
F om he li e a u e e iew, wo ways seem mo e p omising o PHOEBE o inco po a e human beha iou
in a ic simula ion:
phoebe-p ojec .eu
Copy igh © by PHOEBE
134
Deli e able 1.1 – S a e o he A and end-use needs e iew
i) Choice models seems a well-es ablished, lexible and app op ia e me hodology o in oduce
human beha iou as a disc e e inpu on a speci ic en y poin o he a ic simula ion model.
Fo example, a pedes ians’ choice o c oss ou side a c osswalk could be es ima ed as a
p obabili y.
ii) D i e p o iling is easible o be in oduced in he o m o di e en ypes o d i e s (such as
no mal, ina en i e, agg essi e) by changing ce ain simula ion pa ame e s o a p opo ion o
d i e s. Fo example, a ime a e o dis ac ion could be de ined, and he beha iou o he d i e
changed acco dingly.
A p io i isa ion o human isk ac o s is needed, o enable a easible and meaning ul compu a ion o hei
dynamics in a ic simula ion. A e y impo an ques ion is: which human ac o s can be add essed, and by
wha da a? Al hough ideally one would wan o include all he key isk ac o s, i.e. speeding, close- ollowing,
illegal o e aking, a igue, dis ac ion e c., i will be compu a ionally un easible o do so, especially i di e en
oad use ypes a e conce ned, i.e. d i e s, and pedes ians, and cyclis s.
Rega ding cyclis s’ simula ion in pa icula , exis ing ools a e e y basic; e.g. Aimsun has a lane il e ing
module. Howe e , new beha iou al models a e needed o model eliably he ac ual in e ac ions in mode n
u ban se ings, e.g. wi h sha ed space condi ions, e-bikes which a e much as e e c.
O e all, he e a e la ge a ia ions o in as uc u es, a ic, oad use g oups, and he esul ing sa e y
e ec s om hei in e ac ion wi hin ci ies, compa ed o in e u ban a eas – his equi es a sha e o dynamic
inpu s in a ic simula ions.
Bo h oad assessmen and a ic simula ion need o sha pen hei abili y o cap u e beha iou s o VRUs
(pedes ians, cyclis s, e-scoo e s) in a dynamic way, on he basis o new da ase s and me hods.
F om he li e a u e, he ollowing ways seem p omising o PHOEBE o equi ably inco po a e mo o ised and
non-mo o ised oad use s in a ic simula ion and sa e y assessmen , pa icula ly wi h ega d o ajec o y
da a om all oad use ypes – no jus ca s – and in eg a e i wi hin a simula ion ool in o de o iden i y
po en ial con lic s be ween all use s.
VRUs and mic omobili y also need o be included as dis inc a el op ions in o mode choice and modal
shi models.
Mode choice and modal shi a e he i s s eps o simula e a ic on a ne wo k. The e a e wo ypes o
modelling echniques: choice models, and Machine Lea ning models. An impo an ecen de elopmen is
he combined use o bo h echniques, o imp o ed compu a ion. The e is a lo o expe ience in de eloping
such models o u ban a eas, howe e sus ainable modes (o he han public anspo ), mic omobili y and
au oma ed mobili y ha e no been included. Land use and a ic condi ions will ce ainly a ec mode choice,
howe e hei inco po a ion should be imp o ed. Mo eo e , none o he exis ing models conside he
a ailabili y o p io in o ma ion which is la gely a ailable o oad use s, h ough elema ics and social
ne wo ks.
The e a e h ee me hodological di ec ions o he PHOEBE mode choice and modal shi models:
a) Including mic omobil y, VRUs and hei la en a ibu es ia ad anced choice models, enhanced
wi h ML o imp o ed compu a ion.
b) Calcula e modal sha e wi h ine spa ial esolu ion han he ypical o igin-des ina ion le el, h ough
p ope explana o y a iables.
c) Acco dingly, conside “ ou -models” ins ead o “ ip-models”, aking in o accoun ha use s will
swi ch be ween di e en modes along a jou ney, on he basis o dynamic low in o ma ion h ough
apps and social ne wo ks.
Me hodological de ails o hese models will be p ocessed wi hin PHOEBE Task 1.3, howe e i is impo an
a his s age o es ablish hei ole and “en y poin s” in o he o e all me hodological amewo k.
phoebe-p ojec .eu
Copy igh © by PHOEBE
135
Deli e able 1.1 – S a e o he A and end-use needs e iew
5.3 Implica ions on he design o end-use ools and knowledge
p oduc s
The online s akeholde su ey and ocus g oup consul a ions cap u ed he cu en gaps and needs om
he pe spec i e o anspo manage s and municipali ies ( he in ended end use s o he PHOEBE
F amewo k). These a e syn hesised in o a use needs s a emen o in o m he design o he PHOEBE
F amewo k’s end-use ools and knowledge p oduc s.
The online s akeholde su ey was used o unde s and he needs and gaps in cu en p ac ice among he
s akeholde s in ol ed in he PHOEBE use cases (A hens, Valencia, Wes Midlands), as well as a b oade
g oups o anspo manage s eachable h ough he wide ne wo k o con ac s o he p ojec pa ne s.
The su ey pa icipan s indica ed ha he ollowing scena ios need o be p io i ised: i) implemen a ion o
egula o y measu es o limi speeds; ii) in oducing ex ensi e ne wo k o bicycle lanes; iii) p omo ion o
public anspo modes; i ) in oduc ion o new anspo modes; ) implemen ing hie a chical schemes; i)
encou aging modal shi ii) speed calming measu es and iii) expansion o cycling and walking
in as uc u e.
Rega ding he o e all p o essional needs ele an o PHOEBE p ojec , a eas ha we e iden i ied as high
p io i y o esea ch we e:
1) T a ic simula ion en iched wi h in as uc u e sa e y in o ma ion, mode shi and induced demand,
and human beha iou al ac o s
2) Imp o ed accu acy o a ic simula ion, and
3) Road assessmen en iched wi h a ic mic osimula ion in o ma ion and AI/ML models.
When asked abou he use ulness o a po en ial ool o e e yday asks, he su ey pa icipan s placed a
high p io i y on he majo aspec s planned unde he PHOEBE p ojec , namely a ic simula ion en iched
wi h in as uc u e sa e y in o ma ion, modal shi in o ma ion, oad assessmen enhanced wi h AI/ML
models. I is in e es ing o no e ha he esponden s om he PHOEBE use case loca ions in A hens,
Valencia and Wes Midlands e ealed some a iabili y among he needs and p io i ies, suppo ing he
no ion ha he PHOEBE F amewo k needs o able o lexible enough o add ess di e en needs.
The ocus g oup consul a ions d illed mo e deeply in o he cu en uses and p ac ices o oad sa e y
assessmen and mic oscopic simula ion on u ban oad ne wo ks, and he ex en o which hey in o m
decision making in u ban oad planning and managemen .
The esul s o he s akeholde s su ey and he ocus g oups a e summa ized in he o m o a Needs
S a emen on h ee elemen s: i) s a egic goals; ii) decision suppo and daily p ac ice; and iii)
me hodological needs.
The ocus g oups’ pa icipan s aised many ques ions and challenges ha PHOEBE will encoun e . The
eali y o ci ies and public au ho i ies makes i challenging o hem o adop new ways o wo king e en
when he e is he in en ion o. Thus, o a oid alse expec a ions and ensu e mo e e ec i e ou comes, i is
impo an ha PHOEBE ully ecognises i s limi a ions, no as sho comings, bu as le e age as a key playe
in he oad sa e y discussion, ad oca ing o he necessa y changes and new ways o add ess hem.
Alongside PHOEBE’s pilo s, con inuing wi h an hones and open discussion wi h ci ies and o he
s akeholde s is essen ial o ha . The ocus g oups we e an ini ial s ep in his p ocess o acknowledging
hei needs and will con inue h oughou he p ojec in many ways.
phoebe-p ojec .eu
Copy igh © by PHOEBE
136
Deli e able 1.1 – S a e o he A and end-use needs e iew
6 Re e ences
1 Abadi, A, Rajabioun, T & Ioannou, PA (2015), ‘T a ic low p edic ion o oad anspo a ion ne wo ks
wi h limi ed a ic da a,’ IEEE T ansac ions on In elligen T anspo a ion Sys ems, 16(2), pp. 653–662,
doi:10.1109/TITS.2014.2337238.
2 Ada iko u, A., Velaga, N.R. and Mish a, S., 2023. Modelling he e ec o agg essi e d i e beha iou
on longi udinal pe o mance measu es du ing ca - ollowing. T anspo a ion esea ch pa F: a ic
psychology and beha iou , 92, pp.176-200.
3 a Wåhlbe g, A (2008), ‘D i e cele a ion beha iou and acciden s - an analysis,’ Theo e ical Issues in
E gonomics Science, 9, pp. 383–403, doi:10.1080/14639220701596722.
4 a Wåhlbe g, A (2012), ‘Changes in d i e cele a ion beha iou o e ime: Do d i e s lea n om
collisions?’ T anspo a ion Resea ch Pa F: T a ic Psychology and Beha iou , 15(5), pp. 471–479,
doi:h ps://doi.o g/10.1016/j. .2012.04.002.
5 A gha i, A. P., Haque, M. M., Washing on, S., & Smy h, T. (2016). Bayesian la en class sa e y
pe o mance unc ion o iden i ying mo o ehicle c ash black spo s. T anspo a ion esea ch eco d,
2601(1), 90-98.
6 A gha i, A. P., Imani, A. F., Papadimi iou, E., an Gelde , P., & Heza eh, A. M. (2021). Disen angling
he e ec s o unobse ed ac o s on sea bel use choices in mul i-occupan ehicles. Jou nal o choice
modelling, 41, 100324.
7 A gha i, A. P., Ismail, K., Saunie , N., Sha ma, A., & Mi anda-Mo eno, L. (2014, Janua y). Pedes ian-
cyclis in e ac ions a bus s ops along seg ega ed bike pa hs: a case s udy o Mon eal. In P oceedings
o he 93 d Annual Mee ing o he T anspo a ion Resea ch Boa d, Washing on, DC.
8 A gha i, A. P., Papadimi iou, E., Li, X., Kaye, S. A., & O iedo-T espalacios, O. (2021). How much
should a pedes ian be ined o in en ionally blocking a ully au oma ed ehicle? A andom pa ame e s
be a hu dle model wi h he e ogenei y in he a iance o he be a dis ibu ion. Analy ic me hods in
acciden esea ch, 32, 100186.
9 A gha i, A. P., Papadimi iou, E., Pilking on-Cheney, F., Fil ness, A., B ijs, T., B ijs, K., ... & Rod igues,
L. (2022). In es iga ing he e ec s o sleepiness in uck d i e s on hei headway: An ins umen al
a iable model wi h g ouped andom pa ame e s and he e ogenei y in hei means. Analy ic me hods
in acciden esea ch, 36, 100241.
10 A gha i, A. P., Washing on, S., Haque, M. M., & Li, Z. (2018). A comp ehensi e join econome ic model
o mo o ehicle c ashes a ising om mul iple sou ces o isk. Analy ic me hods in acciden esea ch,
18, 1-14.
11 Aghasi, N. H. M. (2019). Applica ion o GIS o U ban T a ic Acciden s: A C i ical Re iew. Jou nal o
Geog aphic In o ma ion Sys em, 11(01), 82–96. h ps://doi.o g/10.4236/jgis.2019.111007
12 Ahmed, H. U., Huang, Y., & Lu, P. (2021). A e iew o ca - ollowing models and modeling ools o
human and au onomous- eady d i ing beha iou s in mic o-simula ion. Sma Ci ies, 4(1), 314-335.
13 Ald ed, R., Bes , L., & Jones, P. (2019). Cyclis s in sha ed bus lanes: could he e be un ecognised
impac s on bus jou ney imes?. In P oceedings o he Ins i u ion o Ci il Enginee s-T anspo (Vol. 172,
No. 3, pp. 135-151). Thomas Tel o d L d.
14 Alexande sson, S., & Johansson, E. (2013). Pedes ians in mic oscopic a ic simula ion. Compa ison
be ween so wa e Viswalk and Legion o Aimsun. Mas e ’s Thesis. Chalme s Uni e si y o
Technology.
15 Ali, M., & Rich e , T. (2022). A F amewo k o Road Sa e y Assessmen and Beha iou al Change.
SSRN Elec onic Jou nal. h ps://doi.o g/10.2139/SSRN.4281920
16 Ali, Y., Haque, M. M., Zheng, Z., & A gha i, A. P. (2022). A Bayesian co ela ed g ouped andom
pa ame e s du a ion model wi h he e ogenei y in he means o unde s anding b aking beha iou in a
connec ed en i onmen . Analy ic me hods in acciden esea ch, 35, 100221.
17 Allen, B., Shin, B., Coope , P. (1978). Analysis o a ic con lic s and collisions. T anspo a ion
Resea ch Reco d 667, 67–72.
phoebe-p ojec .eu
Copy igh © by PHOEBE
143
Deli e able 1.1 – S a e o he A and end-use needs e iew
in T anspo Ne wo ks and Sys ems: 15 h Scien i ic and Technical Con e ence “T anspo Sys ems.
Theo y and P ac ice 2018”, Ka owice, Poland, Sep embe 17–19, 2018, Selec ed Pape s (pp. 53-64).
Sp inge In e na ional Publishing.
138 Giu è, O., G anà, A., Tumminello, M. L., Giu è, T., T ubia, S., S e lazza, A., & Rencelj, M. (2018).
E alua ion o oundabou sa e y pe o mance h ough su oga e sa e y measu es om
mic osimula ion. Jou nal o Ad anced T anspo a ion, 2018, 1-14.
139 Giu è, T., T ubia, S., Canale, A., & Pe saud, B. (2017). Using mic osimula ion o e alua e sa e y and
ope a ional implica ions o newe oundabou layou s o Eu opean Road ne wo ks. Sus ainabili y,
9(11), 2084.
140 Goh, K. C., Cu ie, G., Sa i, M., & Logan, D. (2014). Expe imen al mic osimula ion modeling o oad
sa e y impac s o bus p io i y. T anspo a ion esea ch eco d, 2402(1), 9-18.
141 Gomes, G., May, A., & Ho owi z, R. (2004). Calib a ion o VISSIM o a Conges ed F eeway. Cali o nia
Pa ne s o Ad anced T ansi and Highways (PATH).
142 Gonzalez, MC, Hidalgo, CA & Ba abasi, AL (2008), ‘Unde s anding indi idual human mobili y pa e ns,’
na u e, 453(7196), pp. 779–782.
143 Go ham, R. (2009). Demys i ying induced a el demand.
h ps://ci y2030.o g.ua/si es/de aul / iles/documen s/GIZ_SUTP_TD1_Demys i ying-Induced-T a el-
Demand_EN.pd
144 G ego iades, A. (2007). Towa ds a use -cen ed oad sa e y managemen me hod based on oad a ic
simula ion. In 2007 Win e Simula ion Con e ence (pp. 1905-1914). IEEE.
145 G ego iades, A., Su cli e, A., Papageo giou, G., & Lou ie is, P. (2010). Human-cen e ed sa e y
analysis o p ospec i e oad designs. IEEE T ansac ions on Sys ems, Man, and Cybe ne ics-Pa A:
Sys ems and Humans, 40(2), 236-250.
146 G ie , R., Wickens, C., Kabe , D., S aye , D., Boehm-Da is, D., T a on, J. G., & S . John, M. (2008,
Sep embe ). The ed-line o wo kload: Theo y, esea ch, and design. In P oceedings o he human
ac o s and e gonomics socie y annual mee ing (Vol. 52, No. 18, pp. 1204-1208). Sage CA: Los
Angeles, CA: Sage Publica ions.
147 G oss, F., Pe saud, B. N., & Lyon, C. (2010). A guide o de eloping quali y c ash modi ica ion ac o s
(No. FHWA-SA-10-032). Uni ed S a es. Fede al Highway Adminis a ion. O ice o Sa e y.
148 G uden, C., Campisi, T., Canale, A., Teso ie e, G., & S aml, M. (2019). A c oss-s udy on ideo da a
ga he ing and mic osimula ion echniques o es ima e pedes ian sa e y le el in a con ined space. IOP
Con e ence Se ies: Ma e ials Science and Enginee ing, 603(4). h ps://doi.o g/10.1088/1757-
899X/603/4/042008
149 Guido, G., Gio è, V. P., As a i a, V., & Vi ale, A. (2019, Oc obe ). Using a ic mic osimula ion o
e alua e po en ial c ashes: Some esul s. In 2019 IEEE/ACM 23 d In e na ional Symposium on
Dis ibu ed Simula ion and Real Time Applica ions (DS-RT) (pp. 1-4). IEEE.
150 Guido, G., Saccomanno, F., Vi ale, A., As a i a, V., & Fes a, D. (2011). Compa ing sa e y pe o mance
measu es ob ained om ideo cap u e da a. Jou nal o anspo a ion enginee ing, 137(7), 481-491.
151 Gu ie ez-Oso io, C., & Ped aza, C. (2020). Mode n da a sou ces and echniques o analysis and
o ecas o oad acciden s: A e iew. In Jou nal o T a ic and T anspo a ion Enginee ing (English
Edi ion) (Vol. 7, Issue 4, pp. 432–446). Chang’an Uni e si y. h ps://doi.o g/10.1016/j.j e.2020.05.002
152 Hab emichael, F. G., & de Picado San os, L. (2014). C ash isk e alua ion o agg essi e d i ing on
mo o ways: Mic oscopic a ic simula ion app oach. T anspo a ion esea ch pa F: a ic psychology
and beha iou , 23, 101-112.
153 Hab emichael, F. G., & de Picado San os, L. (2014). C ash isk e alua ion o agg essi e d i ing on
mo o ways: Mic oscopic a ic simula ion app oach. T anspo a ion esea ch pa F: a ic psychology
and beha iou , 23, 101-112.
154 Haddak, M. M., Le è e, M., & Ha e , N. (2016). Willingness- o-pay o oad sa e y imp o emen .
T anspo a ion Resea ch Pa A: Policy and P ac ice, 87, 1–10.
h ps://doi.o g/10.1016/j. a.2016.01.010
155 Hamda , S. (2012). D i e beha iou modeling. In: Eskanda ian, A. (Ed.), Handbook o In elligen
Vehicles. Sp inge , pp. 537–558.
phoebe-p ojec .eu
Copy igh © by PHOEBE
144
Deli e able 1.1 – S a e o he A and end-use needs e iew
156 Hamda , S. H., & Mahmassani, H. S. (2008). F om exis ing acciden - ee ca - ollowing models o
colliding ehicles: explo a ion and assessmen . T anspo a ion esea ch eco d, 2088(1), 45-56.
157 Hamda , S. H., Mahmassani, H. S., & T eibe , M. (2015). F om beha iou al psychology o accele a ion
modeling: Calib a ion, alida ion, and explo a ion o d i e s’ cogni i e and sa e y pa ame e s in a isk-
aking en i onmen . T anspo a ion Resea ch Pa B: Me hodological, 78, 32-53.
158 Hamda , S. H., T eibe , M., Mahmassani, H. S., & Kes ing, A. (2008). Modeling d i e beha iou as
sequen ial isk- aking ask. T anspo a ion esea ch eco d, 2088(1), 208-217.
159 Hansen, J. H., Busso, C., Zheng, Y., & Sa hyana ayana, A. (2017). D i e modeling o de ec ion and
assessmen o d i e dis ac ion: Examples om he UTD i e es bed. IEEE Signal P ocessing
Magazine, 34(4), 130-142.
160 Ha wood, D.; Baue , K.M.; Gilmo e, D.K.; Souley e e, R.; Hans, Z.N. (2010). Valida ion o U.S. Road
assessmen p og am s a a ing p o ocol: Applica ion o sa e y managemen o U.S. oads.
T anspo a ion Resea ch Reco d, 2147, pp. 33-41.
161 Hashmienejad, SHA & Hasheminejad, SMH (2017), ‘T a ic acciden se e i y p edic ion using a no el
mul i-objec i e gene ic algo i hm,’ In e na ional Jou nal o C ashwo hiness, 22(4), pp. 425–440,
doi:10.1080/13588265.2016.1275431.
162 Hayman, T. (2020). How o model cyclis s in Aimsun Nex . Technical no e. URL:
h ps://www.aimsun.com/ echnical-no es/how- o-model-cyclis s-in-aimsun-nex /
163 Haywa d, J. (1971). Nea misses as a measu e o sa e y a u ban in e sec ions. Disse a ion, The
Pennsyl ania S a e Uni e si y.
164 Hegeman, G., Tapani, A., & Hoogendoo n, S. (2009). O e aking assis an assessmen using a ic
simula ion. T anspo a ion esea ch pa C: eme ging echnologies, 17(6), 617-630.
165 Hellmich, S. N. (2017). Wha is Socioeconomics? An O e iew o Theo ies, Me hods, and Themes in
he Field. Fo um o Social Economics, 46(1), 3–25. h ps://doi.o g/10.1080/07360932.2014.999696
166 Helly, W. (1959). Simula ion o bo lenecks in single-lane a ic low. In P oceedings o he Symposium
on Theo y o T a ic Flow
167 Henshe , D. A. ;, & Bu on, K. J. ; (2008). Handbook o anspo modelling: Vol. Volume 1 (D. A.
Henshe & K. J. Bu on, Eds.; Second edi ion). Eme ald. h ps://doi.o g/10.1108/9780857245670
168 Henshe , D. A., & Bu on, K. J. (Eds.). (2007). Handbook o T anspo Modelling (Vol. 1). Eme ald
G oup Publishing Limi ed. h ps://doi.o g/10.1108/9780857245670
169 Henshe , D. A., & G eene, W. H. (2003). The Mixed Logi model: The s a e o p ac ice. T anspo a ion,
30(2), 133–176. h ps://doi.o g/10.1023/A:1022558715350
170 He man, R. (1959). Ca - ollowing and s eady s a e low. In Theo y o T a ic Flow Symposium
P oceedings (pp. 1-13).
171 Hillel, T., Bie lai e, M., Elsha ie, M. Z. E. B., & Jin, Y. (2021). A sys ema ic e iew o machine lea ning
classi ica ion me hodologies o modelling passenge mode choice. Jou nal o Choice Modelling, 38,
100221. h ps://doi.o g/10.1016/j.jocm.2020.100221
172 Holló, P., Eksle , V., & Zukowska, J. (2010). Road sa e y pe o mance indica o s and hei explana o y
alue: A c i ical iew based on he expe ience o Cen al Eu opean coun ies. Sa e y Science, 48(9),
1142–1150. h ps://doi.o g/10.1016/j.ssci.2010.03.002
173 Hol oyd, S. (2021). SLAIN P ojec - Final Repo .
174 Hoogendoo n, R. G., Hoogendoo n, S. P., B ookhuis, K. A., & Daamen, W. (2011). Adap a ion
longi udinal d i ing beha iou , men al wo kload, and psycho-spacing models in og. T anspo a ion
esea ch eco d, 2249(1), 20-28.
175 Hoogendoo n, S. P., Ossen, S., & Sch eude , M. (2006). Empi ics o mul ian icipa i e ca - ollowing
beha iou . T anspo a ion Resea ch Reco d, 1965(1), 112-120.
176 Ho owi z, J. L., Bolduc, D., Di aka , S., Geweke, J., Gonul, F., Haji assiliou, V., Koppelman, F. S.,
Keane, M., Ma zkin, R., Rossi, P., & Ruud, P. (1994). Ad ances in andom u ili y models epo o he
wo kshop on ad ances in andom u ili y models duke in i a ional symposium on choice modeling
beha iou . Ma ke ing Le e s, 5(4), 311–322. h ps://doi.o g/10.1007/BF00999207
177 Ho owi z, J. L., Koppelman, F. S., & Le man, S. R. (1986). A SELF-INSTRUCTING COURSE IN
DISAGGREGATE MODE CHOICE MODELING. FINAL REPORT.
phoebe-p ojec .eu
Copy igh © by PHOEBE
145
Deli e able 1.1 – S a e o he A and end-use needs e iew
178 Hu, L., Ou, J., Huang, J., Wang, F., Wang, Y., Ren, B., ... & Zhou, L. (2022). Sa e y e alua ion o
pedes ian- ehicle in e ac ion a signalized in e sec ions in Changsha, China. Jou nal o T anspo a ion
Sa e y & Secu i y, 14(10), 1750-1775.
179 Huang, F., Liu, P., Yu, H., & Wang, W. (2013). Iden i ying i VISSIM simula ion model and SSAM
p o ide easonable es ima es o ield measu ed a ic con lic s a signalized in e sec ions. Acciden
Analysis & P e en ion, 50, 1014--1024. doi:10.1016/j.aap.2012.08.018
180 Hul k an z, L., Lindbe g, G., & Ande sson, C. (2006). The alue o imp o ed oad sa e y. Jou nal o
Risk and Unce ain y, 32(2), 151–170. h ps://doi.o g/10.1007/s11166-006-8291-z
181 Hup e , C. (1997). Decele a ion o sa e y ime (DST)-a use ul igu e o e alua e a ic sa e y. In ICTCT
Con e ence P oceedings o Semina (Vol. 3, pp. 5-7).
182 Hussein, M., & Sayed, T. (2017). A bi-di ec ional agen -based pedes ian mic oscopic model.
T anspo me ica A: T anspo Science, 13(4), 326-355.
183 Hydén, C. (1987). The de elopmen o a me hod o a ic sa e y e alua ion: The Swedish a ic
con lic s echnique. Lund Uni e si y, Lund, Sweden, Depa men o T a ic Planning and Enginee ing.
184 Hymel, K. M., Small, K. A., & an Dende , K. (2010). Induced demand and ebound e ec s in oad
anspo . T anspo a ion Resea ch Pa B: Me hodological, 44(10), 1220–1241.
h ps://doi.o g/10.1016/j. b.2010.02.007
185 Id, P. M., Con e as, D., & Nica Mo eno, M. (2020). Sa e mobili y, socioeconomic inequali ies, and
aging: A 12-yea mul ile el in e up ed ime-se ies analysis o oad a ic dea h a es in a La in
Ame ican coun y. h ps://doi.o g/10.1371/jou nal.pone.0224545
186 Ilahi, A., Balac, M., Li, A., & Axhausen, Kay. W. (2019). The i s agen -based model o g ea e Jaka a
in eg a ed wi h a mode-choice model. P ocedia Compu e Science, 151, 272–278.
h ps://doi.o g/10.1016/j.p ocs.2019.04.039
187 Imp ialou, M & Quddus, M (2019), ‘C ash da a quali y o oad sa e y esea ch: Cu en s a e and u u e
di ec ions,’ Acciden Analysis P e en ion, 130, pp. 84–90,
doi:h ps://doi.o g/10.1016/j.aap.2017.02.022, oad Sa e y Da a Conside a ions.
188 In e na ional Road Assessmen P og amme (iRAP), 2009. S a Ra ing Roads o Sa e y: The iRAP
Me hodology. iRAP504.04. Si e: www.i ap.o g
189 iRAP (2022), A wo ld ee o high- isk oads. HOW SAFE ARE THE WORLD’S ROADS? WHERE ARE
WE NOW?
190 iRAP. (2014a). iRAP Me hodology Fac Shee # 6 S a Ra ing Sco e equa ions S a Ra ing Sco es.
h ps://i ap.o g/me hodology/
191 iRAP. (2014b). iRAP Me hodology Fac Shee # 6 S a Ra ing Sco e equa ions S a Ra ing Sco es.
h ps://i ap.o g/me hodology/
192 iRAP. (2014c). IRAP Me hodology Fac shee #10 - Casual y es ima ion and calib a ion.
h p://i ap.o g/abou -i ap-3/ esea ch-and- echnical-pape s?download=47: ehicle-speeds-and- he-
i ap-p o ocols.
193 iRAP. (2022). P ocess o aiRAP a ibu e acc edi a ion. h ps://i ap.o g/p ojec /ai- ap/
194 iRAP. (2022a). CycleRAP - Me hodology Fac shee . www.i ap.o g/cycle ap
195 iRAP. (2022b). CycleRAP - Use guide. www.i ap.o g/cycle ap
196 I yo-Asano, M. & Alhajyaseen, W.K.M. (2017). Modeling pedes ian c ossing speed p o iles
conside ing speed change beha iou o he sa e y assessmen o signalized in e sec ions. Accid. Anal.
P e . 108, 332–342
197 Isobe, M., Adachi, T., & Naga ani, T. (2004). Expe imen and simula ion o pedes ian coun e low.
Physica A: S a is ical Mechanics and i s Applica ions, 336(3-4), 638-650.
198 Jabba , R, Shinoy, M, Kha beche, M, Al-Khali a, K, K ichen, M & Ba kaoui, K (2020), ‘U ban a ic
moni o ing and modeling sys em: An io solu ion o enhancing oad sa e y,’ CoRR, abs/2003.07672.
199 Jacobsen, P. (1997, May 14). ‘Liquid’ s. ‘Gas’ Models o T a ic.
200 Jamal, A., Zahid, M., Tauhidu Rahman, M., Al-Ahmadi, H. M., Almoshaogeh, M., Fa ooq, D., & Ahmad,
M. (2021). Inju y se e i y p edic ion o a ic c ashes wi h ensemble machine lea ning echniques: a
compa a i e s udy. In e na ional Jou nal o Inju y Con ol and Sa e y P omo ion, 28(4), 408–427.
h ps://doi.o g/10.1080/17457300.2021.1928233
phoebe-p ojec .eu
Copy igh © by PHOEBE
146
Deli e able 1.1 – S a e o he A and end-use needs e iew
201 Jam oz, K., Romanowska, A., & Budzyński, M. (2018). ANALYSIS OF THE IMPACT OF SOCIO-
ECONOMIC DEVELOPMENT ON ROAD SAFETY BASED ON THE EXAMPLE OF BALTIC SEA
REGION COUNTRIES.
202 Jamson, S., Cho l on, K., & Ca s en, O. (2012). Could in elligen speed adap a ion make o e aking
unsa e?. Acciden Analysis & P e en ion, 48, 29-36.
203 Jas i, V. K., & Higgs, C. F. (2006). A la ice-based cellula au oma a modeling app oach o g anula
low lub ica ion. Jou nal o T ibology, Volume 128, Issue 2. Ap il 2006.
204 Jeong, E., & Oh, C. (2017). E alua ing he e ec i eness o ac i e ehicle sa e y sys ems. Acciden
Analysis & P e en ion, 100, 85-96.
205 Ji, A., & Le inson, D. (2020). A e iew o game heo y models o lane changing. T anspo me ica A:
anspo science, 16(3), 1628-1647.
206 Jian, L. U., & Ying-shuai, L. I. (2017). Re iew and ou look o modeling o lane changing beha iou .
Jou nal o T anspo a ion Sys ems Enginee ing and In o ma ion Technology, 17(4), 48.
207 Jiang, R., Jia, B., & Wu, Q. S. (2004). S ochas ic mul i- alue cellula au oma a models o bicycle low.
Jou nal o Physics A: Ma hema ical and Gene al, 37(6), 2063.
208 Jin S, Qu X, Xu C, e al. (2015). An imp o ed mul i- alue cellula au oma a model o he e ogeneous
bicycle a ic low. Phys Le A; 379: 2409–2416
209 Jo, Y & Ryu, S (2015), ‘Po hole de ec ion sys em using a black-box came a,’ Senso s, 15(11), pp.
29316–29331, doi:10.3390/s151129316.
210 Jo, Y., Jang, J., Pa k, S., & Oh, C. (2021). Connec ed ehicle-based oad sa e y in o ma ion sys em
(CROSS): F amewo k and e alua ion. Acciden Analysis & P e en ion, 151, 105972.
211 Johansson, A. F. (2009). Da a-d i en modeling o pedes ian c owds. D esden: Technische Uni e si a
D esden.
212 Johansson, P.-O. (2001). Is he e a meaning ul de ini ion o he alue o a s a is ical li e? Jou nal o
Heal h Economics, 20(1), 131–139. h ps://doi.o g/10.1016/S0167-6296(00)00073-4
213 Kabe , D. B., Liang, Y., Zhang, Y., Roge s, M. L., & Gangakhedka , S. (2012). D i e pe o mance
e ec s o simul aneous isual and cogni i e dis ac ion and adap a ion beha iou . T anspo a ion
esea ch pa F: a ic psychology and beha iou , 15(5), 491-501.
214 Kahneman, D., & T e sky, A. (2013). P ospec heo y: An analysis o decision unde isk. In Handbook
o he undamen als o inancial decision making: Pa I (pp. 99-127).
215 Kala ian, A., & Fa ooq, B. (2021). Decoding pedes ian and au oma ed ehicle in e ac ions using
imme si e i ual eali y and in e p e able deep lea ning. T anspo a ion esea ch pa C: eme ging
echnologies, 124, 102962.
216 Kalliga, V. (2021). Who Does New T ips and Why?-An Analysis Towa ds he Modelling o Induced
Demand [Mas e hesis]. Technical Uni e si y o Munich.
217 Kamaku a, W., Manski, C. F., & McFadden, D. (1982). S uc u al Analysis o Disc e e Da a wi h
Econome ic Applica ions. Jou nal o Ma ke ing Resea ch, 19(4), 614. h ps://doi.o g/10.2307/3151745
218 Kanugan i, S., Aga wala, R., Du a, B., Bhanegaonka , P. N., Singh, A. P., & Sa ka , A. K. (2017). Road
sa e y analysis using mul i c i e ia app oach: A case s udy in India. T anspo a ion Resea ch P ocedia,
25, 4649–4661. h ps://doi.o g/10.1016/j. p o.2017.05.299
219 Kany, M. S., Ma hiesen, B. V., Sko , I. R., Ko be g, A. D., Thellu sen, J. Z., Lund, H., So knæs, P., &
Chang, M. (2022). Ene gy e icien deca bonisa ion s a egy o he Danish anspo sec o by 2045.
Sma Ene gy, 5, 100063. h ps://doi.o g/10.1016/j.segy.2022.100063
220 Kapa ias, I., Liu, P., Tsaka es os, A., Eden, N., Schmi z, P., Hoadley, S., & Haup mann, S. (2020).
P edic i e oad sa e y impac assessmen o a ic managemen policies and measu es. Case s udies
on anspo policy, 8(2), 508-516.
221 Ka imi, Y., Haghi Kashani, M., Akba i, M., & Mahdipou , E. (2021). Le e aging big da a in sma ci ies:
A sys ema ic e iew. Concu ency and Compu a ion: P ac ice and Expe ience, 33(21).
h ps://doi.o g/10.1002/cpe.6379
222 Ka la is, M. G. (2004). P edic ing Mode Choice h ough Mul i a ia e Recu si e Pa i ioning. Jou nal o
T anspo a ion Enginee ing, 130(2), 245–250. h ps://doi.o g/10.1061/(ASCE)0733-
947X(2004)130:2(245)
phoebe-p ojec .eu
Copy igh © by PHOEBE
147
Deli e able 1.1 – S a e o he A and end-use needs e iew
223 Ka akazas, C., Quddus, M., & Chen, W. H. (2017). A simula ion s udy o p edic ing eal- ime con lic -
p one a ic condi ions. IEEE T ansac ions on In elligen T anspo a ion Sys ems, 19(10), 3196-3207.
224 Kazemzadeh, K., & Bansal, P. (2021). Elec ic bike le el-o -se ice: owa ds he in eg a ion o
hind ance-based and mic osimula ion app oaches. A ailable a SSRN.
225 Kazemzadeh, K., & Sp ei, F. (2022). Towa ds an elec ic scoo e le el o se ice: A e iew and
amewo k. T a el beha iou and socie y, 29, 149-164.
226 Keane, M. P. (1992). A No e on Iden i ica ion in he Mul inomial P obi Model. Jou nal o Business &
Economic S a is ics, 10(2), 193. h ps://doi.o g/10.2307/1391677
227 Khan, G., San iago-Chapa o, K. R., Chi u i, M., & Noyce, D. A. (2010). De elopmen o da a collec ion
and in eg a ion amewo k o oad in en o y da a. T anspo a ion Resea ch Reco d, 2160, 29–39.
h ps://doi.o g/10.3141/2160-04
228 Kha ak, Z. H., Mille , J. S., & Ohlms, P. (2021). Ride-hailing and axi e sus walking: Long e m
o ecas s and implica ions om la ge-scale beha iou al da a. Jou nal o T anspo & Heal h, 22,
101121. h ps://doi.o g/10.1016/j.j h.2021.101121
229 Khel a, B., Ba, I., & To deux, A. (2023). P edic ing highway lane-changing maneu e s: A benchma k
analysis o machine and ensemble lea ning algo i hms. Physica A: S a is ical Mechanics and i s
Applica ions, 128471.
230 Knowles, J., Adams, S., Cue den, R., Sa ill, T., Reid, S., & Tigh , M. (2009). Collisions in ol ing pedal
cyclis s on B i ain’s oads: es ablishing he causes.
h ps:// l.co.uk/uploads/ l/documen s/PPR445_new.pd
231 Kodupugan i, S. R., & Pulugu ha, S. S. (2023). A e Facili ies o Suppo Al e na i e Modes E ec i e
in Reducing Conges ion?: Modeling he E ec o He e ogeneous T a ic Condi ions on Vehicle Delay
a In e sec ions. Mul imodal T anspo a ion, 2(1), 100050.
232 Koenig, J. G. (1980). Indica o s o u ban accessibili y: Theo y and applica ion. T anspo a ion, 9(2),
145–172. h ps://doi.o g/10.1007/BF00167128
233 Koopmann, B., Puch, S., Ehmen, G., & F änzle, M. (2022). Coope a i e maneu e s o highly
au oma ed ehicles a u ban in e sec ions: a game- heo e ic app oach. a Xi p ep in
a Xi :2211.01633.
234 Kouskoulis, G., & An oniou, C. (2017). Sys ema ic e iew o pedes ian simula ion models wi h a ocus
on eme gency si ua ions. T anspo a ion Resea ch Reco d, 2604(1), 111-119.
235 Kouskoulis, G., Spy opoulou, I., & An oniou, C. (2018). Pedes ian simula ion: Theo e ical models s.
da a d i en echniques. In e na ional jou nal o anspo a ion science and echnology, 7(4), 241-253.
236 K onp ase , N., Kuwiboon, P., & Wichi phongsa, W. (2020). Sa e y and ope a ional analysis o median
U- u n in e sec ions in Thailand. GEOMATE Jou nal, 18(68), 156-163.
237 Kumeda, B, Zhang, F, Zhou, F, Hussain, S, Almas i, A & Asse a, M (2019), ‘Classi ica ion o oad a ic
acciden da a using machine lea ning algo i hms,’ in 2019 IEEE 11 h In e na ional Con e ence on
Communica ion So wa e and Ne wo ks (ICCSN), pp. 682–687, doi:10.1109/ICCSN.2019.8905362.
238 Ku ela, B., Das, S., & Dadasho a, B. (2022). Mining pa e ns o au onomous ehicle c ashes in ol ing
ulne able oad use s o unde s and he associa ed ac o s. Acciden Analysis & P e en ion, 165,
106473.
239 la To e, F., Meocci, M., Domenichini, L., B anzi, V., Tanzi, N., & Palio o, A. (2019). De elopmen o
an acciden p edic ion model o I alian eeways. Acciden Analysis and P e en ion, 124, 1–11.
h ps://doi.o g/10.1016/j.aap.2018.12.023
240 La inopoulos, C., Pa ie , A., & Si akuma , A. (2021). Planning o e-scoo e use in me opoli an ci ies:
A case s udy o Pa is. T anspo a ion esea ch pa D: anspo and en i onmen , 100, 103037.
241 Lau eshyn, A., S ensson, Å., & Hydén, C. (2010). E alua ion o a ic sa e y, based on mic o-le el
beha iou al da a: Theo e ical amewo k and i s implemen a ion. Acciden Analysis and P e en ion,
42(6), 1637–1646. h ps://doi.o g/10.1016/j.aap.2010.03.021
242 Lau ina ičius, A., Jukne ičiū ė-Žilinskienė, L., Ra ke ičiū ė, K., Lingy ė, I., Čygai ė, L., G igonis, V.,
Ušpaly ė-Vi kūnienė, R., An o , D., Me s ahi, T., To h-Szabo, Z., & Vá helyi, A. (2012). POLICY
INSTRUMENTS FOR MANAGING ROAD SAFETY ON EU-ROADS. TRANSPORT, 27(4), 397–404.
h ps://doi.o g/10.3846/16484142.2012.751934
phoebe-p ojec .eu
Copy igh © by PHOEBE
148
Deli e able 1.1 – S a e o he A and end-use needs e iew
243 La al, J. A., & Lecle cq, L. (2010). A mechanism o desc ibe he o ma ion and p opaga ion o s op-
and-go wa es in conges ed eeway a ic. Philosophical T ansac ions o he Royal Socie y A:
Ma hema ical, Physical and Enginee ing Sciences, 368(1928), 4519-4541.
244 Lawson, S. (2011). C ash a e–S a Ra ing compa isons. Re iew o a ailable e idence. Wo king Pape
504.2, Eu oRAP, B ussels.
245 Lee, B. H. Y., & Waddell, P. (2010). Residen ial mobili y and loca ion choice: a nes ed logi model wi h
sampling o al e na i es. T anspo a ion, 37(4), 587–601. h ps://doi.o g/10.1007/s11116-010-9270-4
246 Lee, C., So, J., & Ma, J. (2018). E alua ion o coun e measu es o ed ligh unning by a ic simula o –
based su oga e sa e y measu es. T a ic inju y p e en ion, 19(1), 1-8.
247 Lee, D. B., Klein, L. A., & Camus, G. (1999). Induced T a ic and Induced Demand.
H ps://Doi.O g/10.3141/1659-09, 1659, 68–75. h ps://doi.o g/10.3141/1659-09
248 Leona di, S., & Dis e ano, N. (2023). Tu bo-Roundabou s as an Ins umen o Imp o ing he E iciency
and Sa e y in U ban A ea: An I alian Case S udy. Sus ainabili y, 15(4), 3223.
249 Le i, S., Wol , I., Flachsland, C., Koch, N., Kolle , F., & Edmondson, D. (2021). Klimaschu z und
Ve keh : Ziele eichung nu mi unbequemen Maßnahmen möglich. A iadne-Analyse.
h ps://elib.dl .de/147088/
250 Le i a e (2021). Socie al Le el Impac s o Connec ed and Au oma ed Vehicles. Ho izon 2020.
251 Li, L., Li, Y., & Ni, D. (2021). Inco po a ing human ac o s in o LCM using uzzy TCI model.
T anspo me ica B: T anspo Dynamics, 9(1), 198-218.
252 Li, S., Xiang, Q., Ma, Y., Gu, X., & Li, H. (2016). C ash isk p edic ion modeling based on he a ic
con lic echnique and a mic oscopic simula ion o eeway in e change me ging a eas. In e na ional
jou nal o en i onmen al esea ch and public heal h, 13(11), 1157.
253 Li, X., & Sun, J. Q. (2019). Tu ning-lane and signal op imiza ion a in e sec ions wi h mul iple objec i es.
Enginee ing Op imiza ion, 51(3), 484-502.
254 Li, X., Zhang, S., Wu, Y., Wang, Y., & Wang, W. (2021). Explo ing In luencing Fac o s o In e ci y Mode
Choice om View o En i e T a el Chain. Jou nal o Ad anced T anspo a ion, 2021, 1–13.
h ps://doi.o g/10.1155/2021/9454873
255 Linebe ge , R., Hussain, A., Meh a, S., & Pank a z, D. (2018, Janua y 18). Ele a ing he u u e o
mobili y Passenge d ones and lying ca s. Deloi e.
256 Li man, T. (2017). Gene a ed a ic and induced a el. h ps://www. esea chga e.ne /p o ile/Todd-
Li man-
2/publica ion/235360397_Gene a ed_T a ic_and_Induced_T a el_Implica ions_ o _T anspo _Plann
ing/links/5a69 90d4585154d15465728/Gene a ed-T a ic-and-Induced-T a el-Implica ions- o -
T anspo -Planning.pd
257 Liu, M., Zeng, W., Chen, P., & Wu, X. (2017). A mic oscopic simula ion model o pedes ian-pedes ian
and pedes ian- ehicle in e ac ions a c osswalks. PLoS one, 12(7), e0180992.
258 Liu, Q., & Ga be , N. J. (2007). Iden i ying he impac o uck-lane es ic ion s a egies on a ic low
and sa e y using simula ion (No. UVACTS-14-5-103). Uni e si y o Vi ginia. Cen e o T anspo a ion
S udies.
259 Liu, R., da Sil a, J. C., & da Maia Seco, A. J. (2000, July). A bi-modal mic osimula ion ool o he
assessmen o pedes ian delays and a ic managemen . In The P oceedings o he 9 h In e na ional
Associa ion o T a el Beha iou Resea ch Con e ence, Gold Coas , Aus alia.
260 Liu, Y., Dion, F., & Biswas, S. (2006). Sa e y assessmen o in o ma ion delay on pe o mance o
in elligen ehicle con ol sys em. T anspo a ion esea ch eco d, 1944(1), 16-25.
261 Lo enzo Va ela, J. M., Bö jesson, M., & Daly, A. (2018). Public anspo : One mode o se e al?
T anspo a ion Resea ch Pa A: Policy and P ac ice, 113, 137–156.
h ps://doi.o g/10.1016/j. a.2018.03.018
262 Lownes, N. E., & Machemehl, R. B. (2006). Sensi i i y o simula ed capaci y o modi ica ion o VISSIM
d i e beha iou pa ame e s. T anspo a ion Resea ch Reco d, 1988(1), 102-110.
263 Lu, L., Ren, G., Wang, W., Chan, C. Y., & Wang, J. (2016). A cellula au oma on simula ion model o
pedes ian and ehicle in e ac ion beha iou s a unsignalized mid-block c osswalks. Acciden Analysis
& P e en ion, 95, 425-437.
phoebe-p ojec .eu
Copy igh © by PHOEBE
149
Deli e able 1.1 – S a e o he A and end-use needs e iew
264 Lubis, H. al-R., Pan as, V. B., & Fa da, M. (2019). Demand Fo ecas o Jaka a-Su abaya High Speed
Rail based on S a ed P e e ence Me hod. In e na ional Jou nal o Technology, 10(2), 405.
h ps://doi.o g/10.14716/ij ech. 10i2.2442
265 Lucchesi, S. T., E Sil a, J. de A., La anaga, A. M., Zechin, D., & Cybis, H. B. B. (2023). Machine
Lea ning and Image Recogni ion Technologies o Iden i y Buil En i onmen Ba ie s and Incen i es o
Walk. In T anspo a ion Resea ch Reco d (Vol. 2677, Issue 1, pp. 14–24). SAGE Publica ions L d.
h ps://doi.o g/10.1177/03611981221097965
266 Lude e , G., Gün he , C., Sö gel, D., Kos , C., Benke, F., Aue , C., Kolle , F., He bs , A., Rede , K.,
Bö ge , D., Uecke , F., P luge , B., W ede, D., S e le , J., Me o , A., Raune , S., Siala, K., &
Schlichenmaie , S. (2021). Deu schland au dem Weg zu Klimaneu ali ä 2045 - Szena ien und P ade
im Modell e gleich (Zusammen assung). Po sdam Ins i u e o Clima e Impac Resea ch.
h ps://doi.o g/Gunna
267 Lukasik, D., Hale, D., Ma, J., Shibley, P., Malone, T., Chandle , A., Clea y, C., Ma ou , N., & Adebisi,
A. (2020). Enhancing Ac i e T anspo a ion and Demand Managemen (ATDM) wi h Ad anced and
Eme ging Technologies and Da a Sou ces. h ps:// osap.n l.b s.go / iew/do /49721
268 Lundg en, J., & Tapani, A. (2006). E alua ion o sa e y e ec s o d i e assis ance sys ems h ough
a ic simula ion. T anspo a ion esea ch eco d, 1953(1), 81-88.
269 Lynam, D. (2012). De elopmen o Risk Models o he Road Assessmen P og amme. www.i ap.o g.
270 Mahmud, S. S., Fe ei a, L., Hoque, M. S., & Ta assoli, A. (2019). Mic o-simula ion modelling o a ic
sa e y: A e iew and po en ial applica ion o he e ogeneous a ic en i onmen . IATSS esea ch, 43(1),
27-36.
271 Mak, K. K., Sicking, D. L., & Zimme man, K. (1998). Roadside sa e y analysis p og am: A cos -
e ec i eness analysis p ocedu e. T anspo a ion esea ch eco d, 1647(1), 67-74.
272 Manne ing, FL, Shanka , V & Bha , CR (2016), ‘Unobse ed he e ogenei y and he s a is ical analysis
o highway acciden da a,’ Analy ic Me hods in Acciden Resea ch, 11, pp. 1–16,
doi:h ps://doi.o g/10.1016/j.ama .2016.04.001.
273 Ma ensen, H., Diependaele, K., Daniels, S., Van den Be ghe, W., Papadimi iou, E., Yannis, G., Van
Schagen, I., Weije ma s, W., Wijnen, W., Fil ness, A., Talbo , R., Thomas, P., Macha a, K., Aigne
B euss, E., Kaise , S., He mi e, T., Thomson, R., & El ik, R. (2019). The Eu opean oad sa e y
decision suppo sys em on isks and measu es. Acciden Analysis and P e en ion, 125, 344–351.
h ps://doi.o g/10.1016/J.AAP.2018.08.005
274 Maśniak, D. (2007). SOCIAL AND ECONOMIC COSTS OF ROAD ACCIDENTS IN EUROPE.
275 Ma a azzo, TJ, Kondo , D, Mila do, S, Eshke a i, SS, Aan i, P, Pakzad, SN, Buehle , MJ & Ra i, C
(2022), ‘C owdsou cing b idge dynamic moni o ing wi h sma phone ehicle ips,’ Communica ions
Enginee ing, p. 29, doi:10.1038/s44172-022-00025-4.
276 Ma as, K., Bo zo is, G., & Papadopoulos, B. (2021). Sa e y awa e uzzy longi udinal con olle o
au oma ed ehicles. Jou nal o a ic and anspo a ion enginee ing (English edi ion), 8(4), 568-581.
277 Ma as, K., Mak idis, M., Bo zo is, G., Ciu o, B., & Papadopoulos, B. (2019). Fuzzy su oga e sa e y
me ics. In 2019 6 h In e na ional Con e ence on Models and Technologies o In elligen
T anspo a ion Sys ems (MT-ITS) (pp. 1-11). IEEE.
278 McCa hy, L., Delbosc, A., Cu ie, G., & Molloy, A. (2017). Fac o s in luencing a el mode choice
among amilies wi h young child en (aged 0–4): a e iew o he li e a u e. T anspo Re iews, 37(6),
767–781. h ps://doi.o g/10.1080/01441647.2017.1354942
279 McFadden, D. (1974). Condi ional logi analysis o quali a i e choice beha iou . F on ie s in
Econome ics.
280 McFadden, D. (1987). Reg ession-based speci ica ion es s o he mul inomial logi model. Jou nal o
Econome ics, 34(1–2), 63–82. h ps://doi.o g/10.1016/0304-4076(87)90067-4
281 McKenna, F. P., Ho swill, M. S., & Alexande , J. L. (2006). Does an icipa ion aining a ec d i e s' isk
aking?. Jou nal o Expe imen al Psychology: Applied, 12(1), 1.
282 Mclean, R. (2021). Simula ion Modeling o Calga y's E-Scoo e Sys em (Mas e 's hesis, Science).
phoebe-p ojec .eu
Copy igh © by PHOEBE
150
Deli e able 1.1 – S a e o he A and end-use needs e iew
283 Mclean, R., Williamson, C., & Ka an, L. (2021, No embe ). Simula ion Modeling o U ban E-Scoo e
Mobili y. In 2021 29 h In e na ional Symposium on Modeling, Analysis, and Simula ion o Compu e
and Telecommunica ion Sys ems (MASCOTS) (pp. 1-8). IEEE.
284 Mednis, A, Els s, A & Sela o, L (2012), ‘Embedded solu ion o oad condi ion moni o ing using
ehicula senso ne wo ks,’ pp. 1 –5, doi:10.1109/ICAICT.2012.6398502.
285 Meh a, K., Meh an, B., & Hellinga, B. (2019). A me hodology o es ima e he numbe o unsa e ehicle-
cyclis passing e en s on u ban a e ials. Acciden Analysis & P e en ion, 124, 92-103.
286 Michaels, J., Chaumillon, R., Nguyen-T i, D., Wa anabe, D., Hi sch, P., Bella ance, F., ... & Faube ,
J. (2017). D i ing simula o scena ios and measu es o ai h ully e alua e isky d i ing beha iou : A
compa a i e s udy o di e en d i e age g oups. PloS one, 12(10), e0185909.
287 Minde houd, M. & Bo y, P. (2001). Ex ended ime- o-collision measu es o oad a ic sa e y
assessmen . Acciden Analysis and P e en ion 33 (1), 89–97.
288 Miqdady, T., Oña López, R. D., & Oña López, J. D. (2021). Quan i ying he sa e y impac o Connec ed
and Au onomous Vehicles in mo o ways: A simula ion-based s udy. R-E olucionando el anspo e,
2653-2671.
289 Mi a, S., & M. Leon, S. (2014). Disc e e choice model o ai -ca go mode selec ion. The In e na ional
Jou nal o Logis ics Managemen , 25(3), 656–672. h ps://doi.o g/10.1108/IJLM-04-2012-0027
290 Mohammed, H., Bigazzi, A. Y., & Sayed, T. (2019). Cha ac e iza ion o bicycle ollowing and o e aking
maneu e s on cycling pa hs. T anspo a ion esea ch pa C: eme ging echnologies, 98, 139-151.
291 Mohammed, H., Sayed, T., & Bigazzi, A. (2019). Towa d agen -based mic osimula ion o cyclis
ollowing beha iou : Es ima ion o ewa d unc ion pa ame e s using in e se ein o cemen lea ning. In
98 h Annual Mee ing o he anspo a ion esea ch Boa d.
292 Molina, F. E. E., Rami ez, B. D. V. A., Izquie do, F. A., & O ega, D. C. Z. (2021). Road sa e y
pe cep ion ques ionnai e (RSPQ) in La in Ame ica: A de elopmen and alida ion s udy. In e na ional
Jou nal o En i onmen al Resea ch and Public Heal h, 18(5), 1–21.
h ps://doi.o g/10.3390/ije ph18052433
293 Moll, S., López, G., & Ga cía, A. (2021). Analysis o he in luence o spo cyclis s on na ow wo-lane
u al oads using ins umen ed bicycles and mic osimula ion. Sus ainabili y, 13(3), 1235.
294 Mon anino, M., & Punzo, V. (2015). T ajec o y da a econs uc ion and simula ion-based alida ion
agains mac oscopic a ic pa e ns. T anspo a ion Resea ch Pa B: Me hodological, 80, 82-106.
295 Mon ella, A., Guida, C., Mosca, J., Lee, J., & Abdel-A y, M. (2020). Sys emic app oach o imp o e
sa e y o u ban unsignalized in e sec ions: De elopmen and alida ion o a Sa e y Index. Acciden
Analysis & P e en ion, 141, 105523.
296 Mo idpou , S., Sa i, M., & Rose, G. (2010). Lane changing models: a c i ical e iew. T anspo a ion
le e s, 2(3), 157-173.
297 Mo iya, K, Ma sushima, S & Yamanishi, K (2018), ‘T a ic isk mining om he e ogeneous oad
s a is ics,’ IEEE T ansac ions on In elligen T anspo a ion Sys ems, 19(11), pp. 3662–3675,
doi:10.1109/TITS.2018.2856533.
298 Mou akos, V., Oikonomou, M. G., Kopelias, P., Vlahogianni, E. I., & Yannis, G. (2022). Impac s o
au onomous on-demand mobili y se ice: A simula ion expe imen in he Ci y o A hens. T anspo a ion
Le e s, 14(10), 1138-1150.
299 Mpogas, K., Kopelias, P., Mi opoulos, L., & Kepap soglou, K. (2017). Road Sa e y in u ban a eas in
G eece du ing economy down u n. A be o e - A e compa ison. T anspo a ion Resea ch P ocedia,
24, 228–234. h ps://doi.o g/10.1016/j. p o.2017.05.112
300 Muh e , E., & Voll a h, M. (2011). The e ec o isual and cogni i e dis ac ion on d i e ’s an icipa ion
in a simula ed ca ollowing scena io. T anspo a ion esea ch pa F: a ic psychology and beha iou ,
14(6), 555-566.
301 Nan T-W, Mao B-H, Chen Z-J, e al. (2014). U ban mixed nonmo o ehicle low cha ac e simula ion
based on cellula au oma a. J Highw T ansp Res De 2014; 31: 104–109
302 Nazi -Munoz, J. I., Quesnel-Vallée, A., & an den Be g, A. (2021). Global di usion o h ee oad sa e y
policies, 1964–2015. Humani ies and Social Sciences Communica ions 2021 8:1, 8(1), 1–11.
h ps://doi.o g/10.1057/s41599-021-00954-z
phoebe-p ojec .eu
Copy igh © by PHOEBE
151
Deli e able 1.1 – S a e o he A and end-use needs e iew
303 Neilson, A., Ind a mo, Daniel, B., & Tjand a, S. (2019). Sys ema ic Re iew o he Li e a u e on Big Da a
in he T anspo a ion Domain: Concep s and Applica ions. In Big Da a Resea ch (Vol. 17, pp. 35–44).
Else ie Inc. h ps://doi.o g/10.1016/j.bd .2019.03.001
304 Nemas, ASDPSMMCSR, R Kha y (2019), ‘Applica ions o machine lea ning in anspo ,’ ISBN: 978-
1-912433-44-5.
305 Nig o, M., Cas iglione, M., Ma ia Colasan i, F., De Vincen is, R., Valen i, G., Libe o, C., & Comi, A.
(2022). Exploi ing loa ing ca da a o de i e he shi ing po en ial o elec ic mic omobili y.
T anspo a ion Resea ch Pa A: Policy and P ac ice, 157, 78–93.
h ps://doi.o g/10.1016/j. a.2022.01.008
306 Nilsson, R. (2000). D i e s' imp essions o on and ea gaps in queues. E gonomics, 43(12), 1985-
2000.
307 Noland, R. B., & Quddus, M. A. (2006). Flow imp o emen s and ehicle emissions: E ec s o ip
gene a ion and emission con ol echnology. T anspo a ion Resea ch Pa D: T anspo and
En i onmen , 11(1), 1–14. h ps://doi.o g/10.1016/j. d.2005.06.003
308 OECD. (2001). Economic E alua ion o Road T a ic Sa e y Measu es. OECD.
h ps://doi.o g/10.1787/9789282112861-en
309 OECD. (2016). Ze o Road Dea hs and Se ious Inju ies: Leading a pa adigm shi o a sa e sys em.
OECD. h ps://doi.o g/10.1787/9789282108055-en
310 Oikonomou, M. G., O anou, F. P., Vlahogianni, E. I., & Yannis, G. (2020). Impac s o au onomous
shu le se ices on a ic, sa e y and en i onmen o u u e mobili y scena ios. In 2020 IEEE 23 d
In e na ional Con e ence on In elligen T anspo a ion Sys ems (ITSC) (pp. 1-6). IEEE.
311 Oikonomou, M. G., Ziakopoulos, A., Chaudh y, A., Thomas, P., & Yannis, G. (2023). F om con lic s o
c ashes: Simula ing mac oscopic connec ed and au oma ed d i ing ehicle sa e y. Accid. Anal. P e .
187, 107087. h ps://doi.o g/10.1016/j.aap.2023.107087
312 Olmez, S., Douglas-Mann, L., Manley, E., Suchak, K., Heppens all, A., Bi ks, D., & Whipp, A. (2021).
Explo ing he Impac o D i e Adhe ence o Speed Limi s and he In e dependence o Roadside
Collisions in an U ban En i onmen : An Agen -Based Modelling App oach. Applied Sciences, 11(12),
5336.
313 O úza , J. de D., & Willumsen, L. G. (2014). Modelling T anspo . Wiley.
h ps://doi.o g/10.1002/9781119993308
314 Ossen, S., & Hoogendoo n, S. P. (2011). He e ogenei y in ca - ollowing beha iou : Theo y and
empi ics. T anspo a ion esea ch pa C: eme ging echnologies, 19(2), 182-195.
315 O ko ić, I. I., Tollazzi, T., & Š aml, M. (2013). Calib a ion o mic osimula ion a ic model using neu al
ne wo k app oach. Expe sys ems wi h applica ions, 40(15), 5965-5974.
316 Ou ay, F., Galland, S., Gaud, N., & Abbas-Tu ki, A. (2021). Simula ion o connec ed d i ing in
haza dous wea he condi ions: Gene al and ex ensible mul iagen a chi ec u e and models.
Enginee ing applica ions o a i icial in elligence, 104, 104412.
317 O iedo-T espalacios, O., A gha i, A.P. and Haque, M.M., 2020. A hie a chical Bayesian mul i a ia e
o de ed model o dis ac ed d i e s’ decision o ini ia e isk-compensa ing beha iou . Analy ic me hods
in acciden esea ch, 26, p.100121.
318 Ozbay, K., Yang, H., Ba in, B., Mudigonda, S. (2008). De i a ion and alida ion o new simula ion-
based su oga e sa e y measu e. T anspo a ion Resea ch Reco d: Jou nal o he T anspo a ion
Resea ch Boa d 2083, 105–113.
319 Paglia a, F., & P es on, J. (2013). An Induced Demand Model o High Speed 1 in UK. Jou nal o
T anspo a ion Technologies, 03(01), 44–51. h ps://doi.o g/10.4236/j s.2013.31005
320 Pande, A., & Abdel-A y, M. (2006). Assessmen o eeway a ic pa ame e s leading o lane-change
ela ed collisions. Acciden Analysis & P e en ion, 38(5), 936-948.
321 Papadimi iou E., Fa ah H., Van de Kaa G., San oni de Sio F., Hagenzieke M., Van Gelde P.H.A.J.M.
(2022). Towa ds common sa e ‘beha iou ’ s anda ds o au oma ed ehicles. Acciden Analysis &
P e en ion, Volume 174, Sep embe 2022, 106723.
322 Papadimi iou, E., & Yannis, G. (2014). Needs and p io i ies o oad sa e y s akeholde s o e idence-
based policy making. h ps://doi.o g/10.1016/j. anpol.2014.06.006
phoebe-p ojec .eu
Copy igh © by PHOEBE
152
Deli e able 1.1 – S a e o he A and end-use needs e iew
323 Papadimi iou, E., Aube le , J. M., Yannis, G., & Lassa e, S. (2014). Simula ion o Pedes ians and
Mo o ised T a ic: exis ing esea ch and u u e challenges. In e na ional Jou nal o In e disciplina y
Telecommunica ions and Ne wo king (IJITN), 6(1), 57-73.
324 Papadimi iou, E., Lassa e, S., & Yannis, G. (2016). In oducing human ac o s in pedes ian c ossing
beha iou models. T anspo a ion esea ch pa F: a ic psychology and beha iou , 36, 69-82.
325 Papadimi iou, E., Yannis, G., & Golias, J. (2009). A c i ical assessmen o pedes ian beha iou
models. T anspo a ion esea ch pa F: a ic psychology and beha iou , 12(3), 242-255.
326 Papadoulis, A., Quddus, M., & Imp ialou, M. (2019). E alua ing he sa e y impac o connec ed and
au onomous ehicles on mo o ways. Acciden Analysis & P e en ion, 124, 12-22.
327 Pa k, B., & Qi, H. (2006, Sep embe ). Mic oscopic simula ion model calib a ion and alida ion o
eeway wo k zone ne wo k-a case s udy o VISSIM. In 2006 IEEE In elligen T anspo a ion Sys ems
Con e ence (pp. 1471-1476). IEEE.
328 Pa k, S, Sung-min, K & Young-guk, H (2016), ‘Highway a ic acciden p edic ion using ds big da a
analysis,’ The Jou nal o Supe compu ing, 72, pp. 2815–2831.
329 Paschalidis, E., Choudhu y, C. F., & Hess, S. (2019). Combining d i ing simula o and physiological
senso da a in a la en a iable model o inco po a e he e ec o s ess in ca - ollowing beha iou .
Analy ic me hods in acciden esea ch, 22, 100089.
330 Pazzini, M., Cameli, L., Lan ie i, C., Vignali, V., Dondi, G., & Jonsson, T. (2022). New Mic omobili y
Means o T anspo : An Analysis o E-Scoo e Use s’ Beha iou in T ondheim. In e na ional jou nal o
en i onmen al esea ch and public heal h, 19(12), 7374.
331 Pekkanen, J., Giles, O. T., Lee, Y. M., Madigan, R., Daimon, T., Me a , N., & Ma kkula, G. (2021).
Va iable-d i di usion models o pedes ian oad-c ossing decisions. Compu a ional B ain &
Beha iou , 1-21.
332 Pe o , T., Poc a, P., Roman, J., Buzna, Ľ., & Dado, M. (2019). A easibili y s udy o p i acy ensu ing
eme gency ehicle app oaching wa ning sys em. Applied Sciences, 10(1), 298.
333 Phillips, R. O., Ullebe g, P., & Vaa, T. (2011). Me a-analysis o he e ec o oad sa e y campaigns on
acciden s. Acciden Analysis and P e en ion, 43(3), 1204–1218.
h ps://doi.o g/10.1016/j.aap.2011.01.002
334 Pilking on-Cheney, F., A gha i, A. P., Fil ness, A., Papadimi iou, E., Lou enço, A., & B ijs, T. (2021,
June). The i-DREAMS in e en ion s a egies o educe d i e a igue and sleepiness o di e en
anspo modes. In 2021 7 h In e na ional Con e ence on Models and Technologies o In elligen
T anspo a ion Sys ems (MT-ITS) (pp. 1-6). IEEE.
335 Pineda-Ja amillo, J. D. (2019). A e iew o Machine Lea ning (ML) algo i hms used o modeling a el
mode choice. DYNA, 86(211), 32–41. h ps://doi.o g/10.15446/dyna. 86n211.79743
336 Pipes, L. A. (1953). An ope a ional analysis o a ic dynamics. Jou nal o applied physics, 24(3), 274-
281.
337 P ase subpakij, D., & Ni i a ananon, V. (2012). E alua ing accessibili y o Bangkok Me o Sys ems
using mul i-dimensional c i e ia ac oss use g oups. IATSS Resea ch, 36(1), 56–65.
h ps://doi.o g/10.1016/j.ia ss .2012.02.003
338 P ech , L., Keina h, A., & K ems, J. F. (2017a). Iden i ying e ec s o d i ing and seconda y ask
demands, passenge p esence, and d i e cha ac e is ics on d i ing e o s and a ic iola ions–Using
na u alis ic d i ing da a segmen s p eceding bo h sa e y c i ical e en s and ma ched baselines.
T anspo a ion esea ch pa F: a ic psychology and beha iou , 51, 103-144.
339 P ech , L., Keina h, A., & K ems, J. F. (2017b). Iden i ying he main ac o s con ibu ing o d i ing e o s
and a ic iola ions–Resul s om na u alis ic d i ing da a. T anspo a ion esea ch pa F: a ic
psychology and beha iou , 49, 49-92.
340 P es on, A., & Pulugu ha, S. S. (2021). Simula ing and assessing he e ec o a p o ec ed in e sec ion
design o bicyclis s on a ic ope a ional pe o mance and sa e y. T anspo a ion esea ch
in e disciplina y pe spec i es, 9, 100329.
341 P zybyla, J., Taylo , J., Jupe, J. and Zhou, X., 2012, Sep embe . Simpli ied, da a-d i en, e o able ca -
ollowing model o p edic he sa e y e ec s o dis ac ed d i ing. In 2012 15 h In e na ional IEEE
Con e ence on In elligen T anspo a ion Sys ems (pp. 1149-1154). IEEE.
