Fairness on ramp metering. Extending ALINEA for equitable access and congestion reduction

Fai ness on Ramp Me e ing
Ex ending ALINEA o Equi able Access and Conges ion Reduc ion
Yangle Zhan
Supe ision: Ke in Riehl, Anas asios Kou elas, Michail Mak idis
Home Uni e si y Supe ision: Juan Jesús Pé ez
Mas e Thesis Feb ua y 2025
Fai ness in Ramp Me e ing Feb ua y 2025
Fai ness on Ramp Me e ing
Yangle Zhan
IVT
ETH Zü ich
CH-8093 Zu ich
[email p o ec ed]
Supe ision: Ke in Riehl, Anas asios Kou elas,
Michail Mak idis
IVT
ETH Zu ich
Home Uni e si y Supe ision: Juan Jesús Pé ez
Ba celona School o Indus ial Enginee ing
(ETSEIB)
Uni e si a Poli ècnica de Ca alunya
Feb ua y 2025
Abs ac
Ramp me e ing sys ems e ec i ely educe eeway conges ion bu o en ace public op-
posi ion due o inequi able delay dis ibu ion among use s. This s udy add esses his
challenge by in oducing EqALINEA, a ai ness-enhanced ex ension o he ALINEA algo-
i hm designed o balance e iciency and equi y in eeway access. Using mic osimula ions
in SUMO, we e alua e ALINEA, Ups eam ALINEA, and EqALINEA ac oss bo h a
simpli ied ne wo k and a eal-wo ld case s udy o Ba celona’s Ronda de Dal co ido .
Resul s indica e ha adi ional ALINEA p io i izes mainline h oughpu a he expense
o spa ial equi y, disp opo iona ely delaying on- amp use s nea bo lenecks. EqALINEA
mi iga es hese dispa i ies by inco po a ing ai ness cons ain s, including maximum
wai ing h esholds and p oac i e queue managemen . This app oach leads o a mo e
balanced dis ibu ion o delays while main aining o e all a ic e iciency wi hin he
simula ion ime. The algo i hm adap s e ec i ely o u ban en i onmen s such as he
Ronda de Dal , whe e in as uc u e limi a ions and e ol ing mobili y policies equi e
equi able a ic managemen solu ions.
This s udy demons a es he easibili y o in eg a ing ai ness in o decen alized amp
me e ing s a egies o imp o e public accep ance and align wi h sus ainable u ban planning
objec i es.
Keywo ds Keywo ds; Fai ness, Ramp Me e ing, ALINEA, T a ic Equi y, Conges ion
Managemen , U ban Mobili y.
i
Fai ness in Ramp Me e ing Feb ua y 2025
Fai ness on Ramp Me e ing
Yangle Zhan
IVT
ETH Zü ich
CH-8093 Zu ich
[email p o ec ed]
Supe ision: Ke in Riehl, Anas asios Kou elas,
Michail Mak idis
IVT
ETH Zu ich
Home Uni e si y Supe ision: Juan Jesús Pé ez
Ba celona School o Indus ial Enginee ing
(ETSEIB)
Uni e si a Poli ècnica de Ca alunya
Feb ua 2025
Zusammen assung
Zu ah sdosie ung au Au obahnen (Ramp-Me e ing-Sys eme) eduzie en S aus au Au-
obahnen e ek i , s oßen jedoch häu ig au ö en liche Ablehnung au g und ungleiche
Ve zöge ungs e eilung un e den Ve keh s eilnehme n. Diese S udie be ass sich mi
diese He aus o de ung du ch den En wu on EqALINEA, eine ge ech igkei so ien ie en
E wei e ung des ALINEA-Algo i hmus, die E izienz und Ge ech igkei (Fai ness) beim
Au obahnzugang in Einklang b ing . Mi hil e on Mik osimula ionen in SUMO we den
ALINEA, Ups eam ALINEA und EqALINEA sowohl in einem e ein ach en Ne zwe k
als auch in eine ealen Falls udie au de Ronda de Dal in Ba celona e aluie .
Die E gebnisse zeigen, dass de adi ionelle ALINEA-Algo i hmus den Haup ah s ei en-
luss p io isie , dabei jedoch die äumliche Ge ech igkei e nachlässig , wodu ch Nu ze
on Zu ah s ampen, insbesonde e an Engs ellen, übe p opo ional lange Wa ezei en
e leiden. EqALINEA eduzie diese Ungleichhei en du ch die In eg a ion on Fai ness-
Besch änkungen, da un e maximale Wa ezei g enzen und ein p oak i es Wa eschlangen-
managemen . Diese Me hode üh zu eine ausgewogene en Ve eilung de Ve zöge ungen,
wäh end die allgemeine Ve keh se izienz inne halb de Simula ionszei au ech e hal en
bleib . De Algo i hmus pass sich wi ksam an u bane Umgebungen wie die Ronda de
Dal (Ba celona) an, wo in as uk u elle Einsch änkungen und sich wandelnde Mobili ä -
spoli iken ge ech e Ve keh smanagemen lösungen e o de n.
Diese S udie zeig die Machba kei de In eg a ion on Fai ness in dezen alisie e Ramp-
Me e ing-S a egien, um die ö en liche Akzep anz zu e besse n, und mi nachhal ige
S ad planung in Einklang zu b ingen.
Schlüsselwö e Schlüsselwö e ; Ge ech igkei (Fai ness), Au obahn-Zu ah sdosie ung
(Ramp Me e ing), ALINEA, Ve keh sge ech igkei , S aumanagemen , u bane Mobili ä .
ii
Fai ness in Ramp Me e ing Feb ua y 2025
Con en s
Lis o Tables...................................... 3
Lis o Figu es ..................................... 4
Abb e ia ions...................................... 5
1 In oduc ion..................................... 7
1.1 P oblemS a emen .............................. 9
1.2 Mo i a ion................................... 12
1.3 Resea chQues ions.............................. 14
1.4 Objec i es................................... 15
2 Li e a u eRe iew.................................. 16
2.1 Ramp Me e ing Sys ems . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.1 Ramp Me e ing Componen s . . . . . . . . . . . . . . . . . . . . . 17
2.1.2 Ramp Me e ing Con ol S a egies . . . . . . . . . . . . . . . . . 18
2.1.3 Loca ion o Mainline De ec o . . . . . . . . . . . . . . . . . . . . 19
2.1.4 Ca ego ies o Exis ing Ramp Me e ing Algo i hms . . . . . . . . . 21
2.2 Fai ness in Ramp Me e ing . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.1 Fai ness-Conside a ing Ramp Me e ing Algo i hms . . . . . . . . 29
3 Me hodology .................................... 33
3.1 Simula ionModels .............................. 33
3.1.1 ToyModel............................... 34
3.1.2 Ronda de Dal (Ba celona) . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Simula ion Demand Models . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 Toy Model Flow Spawning S a egy . . . . . . . . . . . . . . . . . 42
3.2.2 Ronda de Dal Model Flow Spawning S a egy . . . . . . . . . . . 44
3.2.3 Vehicle & D i e Popula ion . . . . . . . . . . . . . . . . . . . . . 45
3.3 Benchma kCon olle s............................ 49
3.3.1 ALINEA and Ups eam ALINEA . . . . . . . . . . . . . . . . . . 49
3.3.2 Componen Placemen and Da a Collec ion . . . . . . . . . . . . 51
3.3.3 Implemen a ion (ALINEA & Ups eam ALINEA) . . . . . . . . . 53
3.4 EqALINEA .................................. 57
3.4.1 EqALINEA Fai ness . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.5 E alua ionF amewo k ............................ 60
1
Fai ness in Ramp Me e ing Feb ua y 2025
4 Resul sandDiscussion............................... 63
4.1 T a ic E iciency Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2 Fai nessE alua ion.............................. 65
4.2.1 Gini Coe icien Analysis . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.2 On-Ramp Use Expe ience Analysis . . . . . . . . . . . . . . . . . 67
5 Conclusions ..................................... 73
5.1 Limi a ions .................................. 74
5.2 Fu u e Resea ch Di ec ions . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6 Re e ences...................................... 75
A Appendix & Addi ional Da a . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
2

Fai ness in Ramp Me e ing Feb ua y 2025
Lis o Tables
1 Applica ions & P e ious Wo k on Ramp Me e ing . . . . . . . . . . . . . . . . 26
2 O e iew Fai ness-Conside a ing Ramp Me e ing Algo i hms . . . . . . . . . . 31
3 Upda ed Vehicle Type Cha ac e is ics . . . . . . . . . . . . . . . . . . . . . . . 48
4 T a ic E iciency Me ics o Toy Model . . . . . . . . . . . . . . . . . . . . . . 64
5 T a ic E iciency Me ics o Ronda Model . . . . . . . . . . . . . . . . . . . . 64
6 Gini Coe icien s o Toy Model and Ronda De Dal Model . . . . . . . . . . . 65
7 A e age Wai ing Time pe On-Ramp [s] . . . . . . . . . . . . . . . . . . . . . . 68
8 Maximum Wai ing Time pe On-Ramp [s] . . . . . . . . . . . . . . . . . . . . 69
9 Me ging Ra e pe On-Ramp [ eh/cycle] . . . . . . . . . . . . . . . . . . . . . . 70
10 Numbe o Vehicles Joined o Highway [ eh/ST] . . . . . . . . . . . . . . . . . 71
11 Red Time Assigna ion pe On-Ramp (s) . . . . . . . . . . . . . . . . . . . . . 72
12 P esence o Ramp Me e ing Wo ldwide . . . . . . . . . . . . . . . . . . . . . . 81
13 Exi In e al A e age Flow ( eh/h) . . . . . . . . . . . . . . . . . . . . . . . . 81
14 Compa ison o Ramp Me e ing Algo i hms . . . . . . . . . . . . . . . . . . . . 82
3
Fai ness in Ramp Me e ing Feb ua y 2025
Lis o Figu es
1 Ramp Me e ing: How i wo ks. . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2 Ramp Me e ing - Top U.S. Me opoli an A eas. . . . . . . . . . . . . . . . . . 11
3 Global Implemen a ion o Ramp Me e ing (Excluding he Uni ed S a es). . . . 12
4 Compa ison: Mainline Condi ions (wi h and wi hou Ramp Me e ing). . . . . . 16
5 Ramp Me e ing Scheme Fo Gene ic Si e. . . . . . . . . . . . . . . . . . . . . . 18
6 Mainline De ec o Placemen . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
7 ToyModelNe wo k. ................................ 34
8 LesRondesdeBa celona............................... 35
9 T a el Pu pose & Vehicle Occupancy: Dis ibu ion on Ronda de Dal . . . . . . 36
10T a icVolume(Weekday).............................. 36
11 Economic Poles & U ban Cen ali ies. . . . . . . . . . . . . . . . . . . . . . . . 37
12 T a ic Volume Inc ease on Ba celona’s Ring Roads (2021–2023). . . . . . . . . 39
13 Ronda Dal Model Ne wo k. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
14RondaToyModelNe wo k. ............................ 41
15 LOS C i e ia Fo Mul i-Lane Highways. . . . . . . . . . . . . . . . . . . . . . . 42
16 ALINEA Wo king P inciple (Scheme). . . . . . . . . . . . . . . . . . . . . . . . 50
17 Ramp Me e ing Componen Placemen . . . . . . . . . . . . . . . . . . . . . . . 53
18 Ensu ing Su icien Time o On-Ramp F eeway En y. . . . . . . . . . . . . . 57
19 Toy model Occupancy Va ia ion J9. . . . . . . . . . . . . . . . . . . . . . . . . 83
20 Toy model Occupancy Va ia ion J10. . . . . . . . . . . . . . . . . . . . . . . . 84
21 Toy model Occupancy Va ia ion J11. . . . . . . . . . . . . . . . . . . . . . . . 85
22 Toy model A e age Speeds J9. . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
23 Toy model A e age Speeds J10. . . . . . . . . . . . . . . . . . . . . . . . . . . 86
24 Toy model A e age Speeds J11. . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4
Fai ness in Ramp Me e ing Feb ua y 2025
Abb e ia ions
AIMD Addi i e Inc ease Mul iplica i e Dec ease
AD A e age Delay [min/ eh]
ALINEA Asse issemen Linéai e d’En ée Au o ou iè e (Linea eedback amp me e ing
algo i hm)
AMB À ea Me opoli ana de Ba celona (Ba celona Me opoli an A ea)
AR A i al Ra e [%]
ARMS Ad anced Real- ime Ramp Me e ing Sys em
AS A e age Speed [km/h]
DHPC Dual Heu is ic P og amming Con ol
EqALINEA Equi y-o ien ed ALINEA (p oposed ai ness-op imized algo i hm)
GFLC Gene ic Fuzzy Logic Con ol
HERO Heu is ic Ramp-Me e ing Coo dina ion
ILC I e a i e Local Con ol
ITS In elligen T anspo Sys ems
LOS Le el o Se ice
METALINE
Mul i- amp T a ic Managemen Algo i hm o E icien Line In eg a ion and Ne wo k
E iciency
ODMa ix O igin-Des ina ion Ma ix
PCU Passenge Ca Uni
RM Ramp Me e ing
SL2015 SUMO’s 2015 Lane Change Model
5
Fai ness in Ramp Me e ing Feb ua y 2025
ST Simula ion Time
SWARM Sys em Wide Adap i e Ramp Me e ing
SUMO Simula ion o U ban Mobili y ( a ic simula ion so wa e)
TAV To al A i ed Vehicles [ eh
TD To al Delays [h]
TDV To al Depa ed Vehicles [ eh/ST]
TTD To al T a el Dis ance [km]
TTT To al T a el Time [h]
UpALINEA Ups eam ALINEA (modi ied ALINEA wi h ups eam occupancy eedback)
VMS Va iable Message Signs
VSL Va iable Speed Limi s
ZCS Zippe ed Con ol S a egy
6
Fai ness in Ramp Me e ing Feb ua y 2025
can be op imized o balance e iciency wi h ai ness, ensu ing ha he delays imposed on
d i e s a e equi able.
Ramp me e ing aces se e al challenges. S udies highligh a ious obs acles including
issues ela ed o geome y, cos s, public opposi ion, hea y amp olume, and agency
suppo . This is e lec ed in he Uni ed Kingdom’s de elopmen o speci ic guidelines
o amp me e ing sys ems, as shown in documen s like Ramp Me e ing Design Manual
(Cal ans, 2022) and TD 121 (Agency, 2020). These documen s p o ide app aisal and
design equi emen s o amp me e ing. Among hese challenges, 58% o he ba ie s a e
due o he geome y o exis ing in as uc u e (Mizu a e al., 2014b), such as inadequa e
accele a ion leng h, closely spaced amps, and limi ed sigh dis ances. These geome ic
limi a ions make i di icul o ehicles o me ge smoo hly in o mainline a ic, hus
complica ing he implemen a ion o expansion o amp me e ing sys ems.
13

Fai ness in Ramp Me e ing Feb ua y 2025
1.3 Resea ch Ques ions
This hesis explo es how ai ness conside a ions in amp me e ing can be sys ema ically
in eg a ed in o a ic con ol s a egies wi hou signi ican ly comp omising e iciency. To
achie e his, he s udy add esses he ollowing esea ch ques ions:
1. How do exis ing amp me e ing s a egies impac ai ness and e iciency?
•
How do ALINEA and Ups eam ALINEA dis ibu e delays among di e en
on- amp use s?
•Wha equi y conce ns a ise om adi ional amp me e ing s a egies?
2. How can ai ness be explici ly inco po a ed in o amp me e ing con ol?
•
How does EqALINEA modi y he exis ing Local ALINEA amewo k o ensu e
ai e access?
•
Wha mechanisms a e in oduced in EqALINEA o balance ai ness and e i-
ciency?
3. Wha a e he ade-o s be ween ai ness and e iciency in amp me e ing?
•How does EqALINEA compa e o ALINEA and Ups eam ALINEA in e ms
o conges ion delay, wai ing imes on amp, and me ging a e?
•
Does p io i izing ai ness (e.g., educing ex eme wai ing imes) lead o unin-
ended e iciency losses?
4. How do di e en ai ness p inciples eme ge in amp me e ing applica ions?
•
Which ai ness p inciples (Egali a ian, Rawlsian, U ili a ian) a e explici ly o
implici ly applied unde EqALINEA?
•
Does he ai ness concep shi depending on he ne wo k s uc u e (e.g., Toy
Model s. Ronda Toy Model)?
•
How does ai ness a ec di e en use g oups (e.g., local on- amp use s s.
long-dis ance commu e s)?
14
Fai ness in Ramp Me e ing Feb ua y 2025
1.4 Objec i es
A comp ehensi e in es iga ion o he po en ial and equi emen s o implemen ing amp
me e ing s a egies conside ing ai ness is conduc ed in his hesis. The speci ic objec i es
o he p ojec include:
1. E alua e he ai ness o exis ing amp me e ing s a egies
•
Analyze ALINEA and Ups eam ALINEA using ai ness and e iciency me ics,
assessing hei impac on delay dis ibu ion.
•
Compa e delay dispa i ies ac oss di e en on- amps o iden i y po en ial equi y
conce ns in me e ing sys ems.
2.
De elop and implemen EqALINEA, a ai ness-op imized amp me e ing algo i hm
•
Design an enhanced me e ing s a egy ha ensu es equi able access while
p e en ing excessi e conges ion.
•
Implemen EqALINEA in SUMO and alida e i s e ec i eness using mic o-
simula ion models.
3. Analyze he e iciency- ai ness ade-o s in amp me e ing
•
Compa e EqALINEA’s pe o mance agains benchma k con ol s a egy ALINEA
and Ups eam ALINEA in bo h he Toy Model and Ronda Toy Model.
•
Assess key ade-o s be ween h oughpu maximiza ion, conges ion educ ion,
and ai ness imp o emen .
4.
Assess he applicabili y o di e en ai ness p inciples in amp me e ing based on
scena io-speci ic needs.
•
Iden i y which ai ness concep (e.g., Egali a ian, Rawlsian, U ili a ian) is
implici ly o explici ly applied in EqALINEA.
•
E alua e how ai ness in luence equi y ou comes o di e en use g oups (e.g.,
on- amp use s s. main en y use s).
15
Fai ness in Ramp Me e ing Feb ua y 2025
2 Li e a u e Re iew
In his li e a u e e iew, I will i s examine he componen s and con ol s a egies o
amp me e ing o es ablish a ounda ional unde s anding o how hese sys ems wo k. This
will be ollowed by an explo a ion o exis ing amp me e ing algo i hms o ca ego ize
hem and p o ide an o e iew o hei app oaches. Finally, I will e iew algo i hms ha
speci ically inco po a e ai ness conside a ions, ocusing on how hey a emp o add ess
issues o inequi y in delay dis ibu ion.
2.1 Ramp Me e ing Sys ems
Ramp me e s a e a ic signals placed on eeway on- amps o con ol he equency o
ehicles me ging on o he eeway. By managing he low o a ic en e ing he eeway
and dispe sing g oups o ehicles ha complica e me ging, hese signals con ibu e o
educing o e all conges ion. As shown in Figu e 4, ehicles om nea by oads o m a
queue behind he s op line on he amp. They a e hen eleased on o he eeway in lows
o ba ches o ehicles based on he speci ic amp me e ing con ol s a egies in place.
Figu e 4: Compa ison: Mainline Condi ions (wi h and wi hou Ramp Me e ing).
(a) F eeway Wi hou Ramp Me e ing. (b) F eeway Wi h Ramp Me e ing.
Sou ce: Washing on S a e Depa men o T anspo a ion
16
Fai ness in Ramp Me e ing Feb ua y 2025
2.1.1 Ramp Me e ing Componen s
Ramp me e ing has e ol ed signi ican ly om i s ea lies o m, whe e a policeman would
manually di ec a ic en e ing a eeway (Demi al and Celikoglu, 2011). Today, i in ol es
sophis ica ed, in elligen , echnological sys ems designed o op imize a ic low be ween
on- amp ehicles and mainline eeway a ic. These sys ems a e composed o he ollowing
key componen s:
•
Signal Heads: These can be wo-sec ion o h ee-sec ion signals. Two-sec ion
heads ea u e g een and ed ligh s, while h ee-sec ion heads include a yellow ligh ,
p o iding mo e amilia cues o d i e s. The signal heads manage when ehicles can
p oceed on o he eeway.
•
De ec o s: Senso equipmen (loop de ec o s) moni o s a ic condi ions bo h on
he amps and he mainline. Ramp de ec o s ensu e ha he signal only u ns g een
when a ehicle is p esen a he s op line. Queue de ec o s can moni o he leng h o
he line wai ing o en e he eeway, while o he de ec o s assess a ic low on he
eeway i sel o de e mine he op imal me e ing a e. These de ec o s a e ypically
connec ed di ec ly o he amp con olle .
•
Signage: Signs play a c ucial ole in guiding d i e s. They should be placed a he
s a o he amp and nea he signal o p o ide clea ins uc ions. Addi ional signs
ups eam o wi h adap i e sc eens can ale d i e s i he amp is being ac i ely
me e ed, helping o manage expec a ions and imp o e compliance wi h he sys em.
Toge he , hese componen s ensu e smoo h and e icien amp me e ing, minimizing
conges ion and imp o ing eeway sa e y.
17
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 5: Ramp Me e ing Scheme Fo Gene ic Si e.
Sou ce: Highways Agency (2008)
2.1.2 Ramp Me e ing Con ol S a egies
RM employs a ious con ol s a egies o manage a ic e ec i ely. Mainly h ee ypes o
Ramp Me e ing Con ol sys ems can be dis inguished:
•
Local sys em: This app oach makes amp me e ing decisions independen ly o
each amp, conside ing only he a ic condi ions a he speci ic amp and i s
immedia e eeway en y poin , wi hou aking in o accoun he a ic condi ions
ac oss he en i e ne wo k.
•
Sys em-wide con ol / Coo dina ed sys em: This app oach coo dina es amp
me e s ac oss mul iple on- amps and he en i e eeway sys em. I adop s a holis ic
s a egy by using da a om a ious amps and eeway sec ions o manage a ic
low ac oss a la ge a ea. Al hough mo e complex, i can be e dis ibu e a ic
18

Fai ness in Ramp Me e ing Feb ua y 2025
and educe conges ion on he main oad.
•
In eg a ed sys em: This app oach ex ends beyond amp me e ing by in eg a ing
a ious a ic con ol measu es, such as signal iming, amp me e ing, and ou e
guidance h ough Va iable Message Signs (VMS). The goal is o op imize o e all
a ic low by u ilizing a ange o ools and s a egies o manage eeway a ic
mo e e ec i ely.
Based on he con ol philosophy, he e a e wo ca ego ies o con ol schemes:
•
P e- imed con ol: This is he simples con ol s a egy, bu i p o ides he leas
e ec i e esul s. I elies on p e-de e mined, s a ic iming o he amp signal, wi h
he me e ing a e se based on his o ical a ic pa e ns a he han eal- ime da a.
I ’s simple bu less lexible, as i doesn’ adjus o cu en a ic condi ions.
•
T a ic- esponsi e con ol: This s a egy dynamically adjus s he me e ing a e
based on eal- ime a ic condi ions. By u ilizing de ec o s o moni o a ic low,
speed, and conges ion bo h on he amp and he eeway, he sys em can espond o
luc ua ions and op imize a ic low mo e e ec i ely.
2.1.3 Loca ion o Mainline De ec o
Ramp me e ing sys ems ely on s a egically placed de ec o s o moni o a ic condi ions
and in o m con ol decisions. A c i ical design choice is he placemen o mainline de ec o s,
which a e ypically ca ego ized as ups eam o downs eam ela i e o he me ging a ea
(Figu e 6).
•
Ups eam de ec o s a e posi ioned be o e he me ging zone o moni o app oaching
eeway a ic. These de ec o s enable p oac i e con ol by p edic ing conges ion
p opaga ion and adjus ing me e ing a es p eemp i ely.
•
Downs eam de ec o s a e placed a e he me ging a ea o di ec ly obse e
conges ion le els a he bo leneck. This eac i e app oach p io i izes s abilizing
downs eam a ic low, as seen in s a egies like ALINEA, which uses downs eam
occupancy o egula e amp in lows.
19
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 6: Mainline De ec o Placemen .
Sou ce: an Lindonk (2020)
The choice be ween ups eam and downs eam de ec o placemen signi ican ly impac s
algo i hm beha io because:
1. Ups eam de ec o s:
•Allow p oac i e con ol, an icipa ing conges ion be o e i occu s.
•Enable be e queue managemen on he amp.
•P o ide mo e immedia e esponse o changes in amp demand.
2. Downs eam de ec o s:
•Measu e he ac ual impac o me ging a ic on mainline low
•Can de ec conges ion o ma ion mo e di ec ly
•Allow o eedback con ol based on he esul ing a ic s a e a e me ging
This undamen al dis inc ion explains why di e en algo i hms adop speci ic de ec o
con igu a ions, as sys ema ically compa ed in Table 14 by Luaibi e al. (2023). The able
highligh s how ups eam de ec o -based s a egies p io i ize conges ion p e en ion while
downs eam- ocused app oaches excel a bo leneck s abiliza ion.
Ups eam placemen ends o a o esponsi e local con ol, while downs eam placemen
is o en used in algo i hms aiming o op imal mainline low condi ions. The choice
in luences he algo i hm’s esponsi eness, con ol s a egy, and o e all e ec i eness in
managing a ic low.
20
Fai ness in Ramp Me e ing Feb ua y 2025
2.1.4 Ca ego ies o Exis ing Ramp Me e ing Algo i hms
Ramp me e ing algo i hms can be ca ego ized in o ou di e en ypes (Zhang e al., 2001)
based on he combina ion o con ol s a egies and he deg ee o which hey inco po a e
local and sys em-wide a ic condi ions.
1.
Isola ed algo i hms: Each on- amp ope a es independen ly, meaning ha i s
me e ing a e is de e mined based only on he a ic condi ions a ha speci ic
amp. These condi ions can include ac o s like ehicle low, occupancy, a el speed,
and po en ially any queue o e low ha builds up on he amp i sel .
•
Zone Algo i hm
2
:This algo i hm egula es eeway a ic by di iding he
oadway in o zones and assigning each a calcula ed me e ing a e o main ain
he a ge densi y. The assigned a e is dis ibu ed among he amps based on
p ede e mined ac o s, adjus ing en y low o minimize conges ion.
•
ALINEA
3
:The ALINEA algo i hm adjus s he me e ing a e using occupancy
da a collec ed om a downs eam de ec o on he eeway. E e y ew seconds,
he algo i hm calcula es he di e ence be ween he cu en occupancy and a
a ge h eshold. I hen upda es he me e ing a e o educe his di e ence,
aiming o keep eeway a ic below a conges ion h eshold.
•
Local me e ing using neu al ne wo ks: This adap i e amp me e ing
app oach le e ages neu al ne wo k models o de e mine me e ing a es o
indi idual amps. Unlike adi ional algo i hms ha ely on p e-se h esholds
o simple eedback mechanisms, neu al ne wo ks can lea n complex pa e ns in
a ic low by analyzing his o ical and eal- ime da a. This enables he sys em
o p edic op imal me e ing a es based on a a ie y o a ic condi ions, such
as amp occupancy, mainline a ic densi y, and speed.
2.
Coope a i e algo i hms: These algo i hms coo dina e he me e ing a es o
mul iple on- amps along a eeway. Unlike isola ed amp me e ing, whe e each
amp ope a es independen ly, coope a i e algo i hms u ilize da a om neighbo ing
amps and o e all eeway condi ions o collabo a i ely adjus me e ing a es. This
app oach helps p e en bo h bo leneck conges ion and spillback a c i ical amps,
op imizing a ic low no only a indi idual amps bu ac oss he en i e eeway.
2
S a i ied Zone Me e ing: Va ian o Zone me e ing, ha using eal- ime densi y da a o mo e p ecise
con ol.
3
Ex ension o ALINEA: Sma agdis e al. (2004) p oposed h ee mo e modi ica ions FL-ALINEA,
UP-ALINEA and X-ALINEA/Q o imp o ing he pe o mance in a ious scena ios.
21
Fai ness in Ramp Me e ing Feb ua y 2025
•
Helpe amp algo i hm: This me hod in eg a es local a ic- esponsi e
con ol wi h a cen al o e ide mechanism o achie e coo dina ed, sys em-wide
managemen . On- amps a e o ganized in o g oups, each assigned p ede ined
me e ing a es ha dynamically adjus based on local a ic condi ions. I
he queue a a pa icula amp su passes a se h eshold, he cen al o e ide
ac i a es o adjus he me e ing a e, aiming o p e en excessi e queue buildup
and conges ion spillback. When he issue pe sis s, he algo i hm applies mo e
es ic i e me e ing a es ups eam o u he alle ia e conges ion.
•
Linked- amp algo i hm: Each amp’s me e ing a e is de e mined by
sub ac ing he ups eam a ic low om a p ede ined a ge low a e. When
one amp’s me e ing a e eaches i s minimum le el due o conges ion, ups eam
amps a e also equi ed o educe hei me e ing a es o main ain coo dina ed
low.
3.
Compe i i e algo i hms: These algo i hms calcula e wo se s o me e ing a es o
each amp: one based on local a ic condi ions (e.g., a ic low, speed, occupancy
a he amp) and ano he based on sys em-wide o co ido -le el condi ions (e.g.,
conges ion le els o low on adjacen amps and he mainline). The mo e es ic i e
o hese wo a es is hen selec ed and implemen ed a each amp.
•
Compass algo i hm: This algo i hm adjus s on- amp me e ing a es using
eal- ime eeway da a and p ede ined h esholds o occupancy and a ic
olume. By moni o ing local and downs eam mainline condi ions, Compass
selec s he mos es ic i e me e ing a e o p e en conges ion, while also
employing queue spillback de ec ion o ensu e ha amp queues do no dis up
adjacen su ace oads. This sys em can ope a e in bo h au oma ed and manual
modes, allowing a ic manage s lexibili y o o e ide au oma ed se ings as
needed o inciden esponse o special condi ions.
•
Bo leneck algo i hm: This algo i hm is speci ically designed o manage
a ic low nea conges ion-p one a eas, o "bo lenecks," on a eeway. I
ope a es by balancing he incoming a ic demand ups eam o a amp wi h
he capaci y a ailable a he downs eam bo leneck, adjus ing he me e ing
a e a on- amps acco dingly. When conges ion begins o o m, he algo i hm
educes me e ing a es o all amps in he a ec ed zone, coo dina ing en y
a es o p e en an o e load a he bo leneck. Addi ionally, i inco po a es a
queue managemen sys em ha adjus s me e ing a es i amp queues exceed
a se h eshold, p e en ing spillback on o su ace s ee s.
22
Fai ness in Ramp Me e ing Feb ua y 2025
Egali a ian ai ness in amp me e ing would seek o p o ide equal wai imes o all
d i e s, ensu ing ha no one amp o use is p io i ized o e ano he . This app oach
would s anda dize me e ing a es ac oss all amps so ha all d i e s expe ience
he same le el o access, ega dless o hei loca ion o ehicle ype. This p inciple
sac i ices some e iciency in a o o s ic equali y, dis ibu ing wai imes uni o mly
ac oss use s.
Dio ho ic ai ness e e s o ai ness in sys ems whe e indi iduals ac i ely in e ac and
in luence esou ce alloca ion h ough mechanisms such as ma ke s o p icing. In his
amewo k, use s make decisions based on hei p e e ences and cons ain s, such as budge s
o willingness o pay, wi h ou comes eme ging om hese in e ac ions. Examples include
conges ion p icing o oll sys ems, whe e d i e s choose whe he o pay o p io i ized
access. Howe e , in his s udy he scope does no include hose ac o s and p ima y
ope a es wi hin a dianeme ic ai ness amewo k. So, his s udy ocuses exclusi ely on he
dianeme ic ai ness aspec s o esou ce alloca ion in a ic managemen .
2.2.1 Fai ness-Conside a ing Ramp Me e ing Algo i hms
The ade-o be ween e iciency and ai ness in eeway amp me e ing has been a ecu ing
heme in he li e a u e (Ko sialos and Papageo giou (2001); Le inson e al. (2002)). As
obse ed in hese s udies, he mos e icien amp me e ing s a egies o en achie e hei
objec i es a he cos o equi y, c ea ing a undamen al ension in amp con ol. The
mos e icien amp con ol logic ends o p io i ize ee- low condi ions on he eeway
mainline, bene i ing he majo i y o commu e s while hea ily es ic ing access a speci ic
on- amps. This app oach minimizes conges ion bu concen a es delays a ce ain en y
poin s, making i less equi able. While his s a egy may sa is y enginee ing s anda ds o
e iciency, i isks low public accep ance due o pe cei ed un ai ness.
To mi iga e hese issues, coo dina ion o on- amp me e s is equen ly employed, no
only o p e en eeway queues bu also o dis ibu e delays mo e e enly. Al hough
o mal heo e ical models o equi y in amp con ol a e limi ed, p ac ical measu es aimed
a enhancing ai ness ha e e ol ed o e ime in eal-wo ld applica ions. Fo ins ance,
maximum queue leng h cons ain s, o iginally in ended o p e en spillo e s on o local
s ee s, also se e o limi excessi e wai imes o amp use s. Many s a egies u he
inco po a e minimum and maximum me e ing a e limi s o achie e a mo e balanced
29

Fai ness in Ramp Me e ing Feb ua y 2025
dis ibu ion o delays. Some sys ems, like Den e ’s “helpe algo i hm,” Lipp e al. (1991)
adjus ups eam me e ing a es o elie e conges ion on hea ily es ic ed downs eam
amps. Simila ly, he Minneso a Zonal and S a i ied Zonal algo i hms impose maximum
delay limi s o ensu e ha no d i e expe iences disp opo iona ely long wai s a a single
amp.
The ollowing Table 2 p o ides a compa ison o a ious s udies ha examine he aspec o
ai ness in amp me e ing algo i hms.
30
Fai ness in Ramp Me e ing Feb ua y 2025
Table 2: O e iew Fai ness-Conside a ing Ramp Me e ing Algo i hms
Name S udy Desc ip ion and Key Fea u es Fai ness P inciple E iciency s. Fai ness T ade-O S eng hs Limi a ions Pa ame e s
Balancing E iciency
and
Equi y o Ramp Me e s
Zhang and Le inson (2005)
BEEX is a amp me e ing algo i hm ha minimizes weigh ed a el ime,
assigning highe weigh s o longe wai imes a on- amps o balance
e iciency wi h equi y. I uses an adjus able equi y coo dina ion ac o
o dis ibu e delays mo e e enly ac oss amps. Real- ime da a is used o
iden i y bo lenecks and adjus amp me e ing a es.
U ili a ian (F eeway d i e s)
wi h
Ha sanyian (On- amp d i e s)
-Allows a unable balance be ween e iciency and
equi y by adjus ing he equi y coo dina ion ac o .
-Highe alues inc ease ai ness bu may educe e iciency.
-P io i izes eeway mainline e iciency,
keeping i a o below capaci y.
-Possible spillo e on o local s ee s
-Does no p oac i ely add ess equi y
o on- amp use s.
-A e age On-Ramp
Wai ing Time
A Pa e o-Op imiza ion
App oach o a
Fai Ramp Me e ing
Li e al. (2016)
This algo i hm achie es ai amp me e ing by minimizing bo h o al
sys em delay ( eeway and on- amp) and spa ial inequi y in delay
dis ibu ion ac oss on- amps. I uses a Pa e o-op imal mul i-objec i e
op imiza ion app oach, wi h a spa ial equi y index o compa e delays,
dynamically inding solu ions based on desi ed equi y-e iciency balance.
Egali a ian (On- amp d i e s)
wi h
Ha sanyian (F eeway d i e s)
-Balances sys em e iciency (minimizing o al delay)
wi h equi y by gene a ing a Pa e o on o solu ions,
allowing ade-o s be ween objec i es.
-Ensu es ai e delay dis ibu ion ac oss
amp g oups
-P o ides Pa e o-op imal solu ions o
decision-make s.
-P io i izing equi y may educe
eeway e iciency.
-Spa ial equi y index does no
accoun o empo al equi y.
-To al T a el Delay
( eh·h) and Equi y Index
An Analysis on E iciency
and Equi y o
Fixed-Time Ramp Me e ing
Kes en e al. (2013)
FTRM uses ixed signal imings based on his o ical a ic da a o con ol
on- amp in lows and manage eeway capaci y, wi hou eal- ime
adjus men s. The s a egy was es ed in VISSIM on a conges ed co ido ,
ocusing on impac s o delays and speeds.
Fai ness was assessed by a iabili y: lowe a iabili y in speed o delay
means a ai e sys em. The s udy shows ha applying FTRM makes
condi ions less ai o on- amp d i e s.
Egali a ian
(equal ea men o all ehicles,
ega dless o needs o loca ions)
-E iciency gains on he mainline may come a he expense
o inc eased delays a on- amps;
-Equi y ou come depends on speci ic measu es used
(e.g., delay s. speed).
-Simple o implemen and cos -e ec i e
compa ed o dynamic me e ing;
-Imp o es mainline e iciency.
-Possible spillo e on o local s ee s;
-Lacks eal- ime adap abili y;
-May wo sen equi y o amp use s.
-Spo Speed
-Space Mean Speed
-Delay
Coo dina ed Ramp Me e ing
wi h Equi y Conside a ion
Using Rein o cemen Lea ning
Lu e al. (2017)
RAS-EQ uses ein o cemen lea ning (RL) o op imize amp me e ing,
balancing a ic e iciency and use equi y by dynamically adjus ing
me e ing a es based on e iciency ( o al ime spen ) and equi y
(e en delay dis ibu ion).
U ilizes asymme ic cell ansmission model (ACTM) o a ic low
and adjus s based on e iciency and equi y ewa d unc ions.
U ili a ian (F eeway d i e s)
wi h
Egali a ian (On- amp d i e s)
-Allows adjus men ia pa ame e δ o weigh e iciency
and equi y; highe δ alues imp o e equi y a he cos o
some e iciency.
-Highly e icien o eeway low by
p io i izing mainline e iciency.
-Be e ensu e ai e dis ibu ion o
delays among g oups o amps.
-SD o TWT does no accoun o
empo al equi y among ehicles
a i ing a di e en imes.
-Requi es signi ican compu a ional
esou ces o aining.
-SD o TWT
-To al Time Spen (TTS)
E iciency and Equi y
Pe o mance
o a Coo dina ed
Ramp Me e ing Algo i hm
Li e al. (2016)
This algo i hm modi ies HERO o inco po a e a combined index o
e iciency ( o al a el ime) and equi y (Gini coe icien )
o balance objec i es.
The Gini coe icien measu es spa ial equi y o ensu e balanced delay
dis ibu ion ac oss amps, es ed in AIMSUN simula ion.
U ili a ian (g ea e -good ocus)
wi h Egali a ian elemen s
(using Gini o spa ial equi y)
-Allows unable adjus men s o p io i ize ei he e iciency
o equi y;
-Gini coe icien p o ides balanced delay dis ibu ion.
-Ensu es equi able delay dis ibu ion
ac oss amps, main aining mainline
e iciency.
-May sligh ly educe e iciency
i equi y objec i e is weigh ed hea ily;
-Focuses on spa ial equi y wi hou
empo al conside a ions.
-To al T a el Time (TTT),
-Gini Coe icien
New Ho izon al Equi y
Measu e o Ramp Me e s Amini e al. (2016)
A ho izon al equi y measu e implemen ed wi h HERO, assigning delays
based on indi idual con ibu ions o conges ion using ehicle O-D da a.
Delays a e p opo iona ely assigned acco ding o each ehicle’s impac
on downs eam bo lenecks, es ed in hypo he ical mic osimula ion.
A is o elian (p opo ional delays
based on conges ion con ibu ion)
-Enhances eeway low and ai ness by delaying
high-impac ehicles, hough i may inc ease me e ing a es
o amps wi h high downs eam impac .
-P o ides mo e equi able delays by
penalizing high-impac amps,
po en ially imp o ing public accep ance.
-Requi es eal- ime O-D da a;
-Low-impac ehicles may ace delays
i queued behind high-impac ehicles.
-No malized Gini coe icien
E iciency e sus Fai ness in
Ne wo k-Wide Ramp Me e ing Ko sialos and Papageo giou (2001)
AMOC dynamically adjus s me e ing a es based on a ic demands
and amp s o age, balancing eeway e iciency wi h amp equi y.
Real- ime da a is used o adjus men s, and queue cons ain s p e en
excessi e delays, es ed on he Ams e dam ing- oad.
U ili a ian (F eeway d i e s)
wi h
Ha sanyian (On- amp d i e s)
-P io i izes mainline e iciency by ocusing me e ing a
amps nea bo lenecks. When queue cons ain s a e ac i e,
AMOC sac i ices some eeway h oughpu o educe
excessi e amp delays and a oid spillo e .
-Maximizes eeway low while applying
queue cons ain s o ai e delay dis ibu ion.
-Fai ness is seconda y o e iciency,
lacking a speci ic ai ness objec i e
unc ion, which may esul in unequal
ea men o di e en amps.
-Spa ial equi y (Queue Leng h)
Measu ing he equi y
and e iciency o amp me e s Le inson e al. (2004)
Co-EOAX minimizes o al weigh ed a el ime by applying non-linea
weigh s o amp delays, balancing e iciency and equi y wi hou
equi ing O-D da a. The Global On-Ramp G ouping Fac o X enables
lexibili y in adjus ing equi y, es ed in AIMSUN.
U ili a ian (F eeway d i e s)
wi h
Egali a ian (On- amp d i e s)
Highe X alues imp o e equi y by dis ibu ing delays mo e
e enly ac oss amps bu may educe eeway h oughpu .
-Simpli ies equi y adjus men s wi hou
O-D da a, acili a ing easie implemen a ion.
-Limi ed by p e-se X alues;
-Highe equi y se ings may lowe
mainline e iciency.
-To al Weigh ed T a el Time (WTT)
-G oup Fac o X
Sou ce: Sel -compiled da a.
31
Fai ness in Ramp Me e ing Feb ua y 2025
The concep o equi y in amp me e ing is undamen al o gaining public accep ance and
ensu ing success ul implemen a ion. Ramp me e ing inhe en ly in ol es ade-o s- some
d i e s bene i om imp o ed eeway low, while o he s, pa icula ly hose delayed a
on- amps, may eel disad an aged o e en ha med by inc eased wai ing imes. This aises
an impo an ques ion: Wha is ai ness, who bene i s and who is disad an aged, and
how much, compa ed o a sys em wi hou amp me e ing? Fu he mo e, i he sys em
sac i ices he equi y o on- amp d i e s o imp o e he o e all co ido e iciency, Wha is
he equi y-e iciency ade-o a io in amp me e ing, speci ically quan i ying he deg ee o
delay imposed on on- amp d i e s o achie e imp o emen s in o e all co ido e iciency?
F eeways we e ini ially buil o in e ci y a el, bu o e ime, hey ha e e ol ed o
mee he needs o local commu e s as well. This shi p omp s a ee alua ion o whe he
long-dis ance a el should emain a p ima y policy ocus, o i i would be jus i iable o
educe some eeway e iciency o enhance access o on- amp use s, pa icula ly in a eas
nea o bo lenecks. As we can obse e om p e ious esea ch, s udies ha e inc easingly
shi ed ocus om solely maximizing eeway e iciency (u ili a ian ai ness) o ecognizing
and add essing he equi y o on- amp d i e s (egali a ian ai ness), e en i i means
comp omising e iciency o some ex en .
The majo i y o hose s udies use coo dina ed, a ic- esponsi e amp me e ing, which
is no only complex bu also cos ly o implemen , and i does no di ec ly add ess use s’
equi y conce ns.
Fo his eason, his esea ch aims o explo e whe he amp-me e ed sys ems genuinely
bene i o disad an age hose delayed a amps, pa icula ly use s joining nea bo lenecks.
To do his, bo h he ALINEA algo i hm and a modi ied e sion o ALINEA will be
implemen ed in a eal-case simula ion using SUMO. This app oach will assess he angible
impac s on use expe ience, explo ing bo h he e iciency and ai ness o amp me e ing
om a use -cen e ed pe spec i e.
32
Fai ness in Ramp Me e ing Feb ua y 2025
3 Me hodology
This chap e desc ibes he me hodology used in his hesis, which consis s o de eloping
simula ion models, implemen ing benchma k con olle s, designing an op imized amp
me e ing algo i hm, and e alua ing i s pe o mance.
To sys ema ically analyze amp me e ing s a egies, he s udy employs mic oscopic a ic
simula ion models, which allow o de ailed ehicle-le el in e ac ions. These simula ions
help in e alua ing bo h ai ness and e iciency o di e en amp me e ing echniques.
The ollowing sec ions p o ide an o e iew o he simula ion models used, he bench-
ma k con olle s implemen ed, he p oposed op imiza ion app oach, and he e alua ion
amewo k:
•
Simula ion Models: This sec ion desc ibes he oy model, which se es as a
simpli ied es en i onmen , and he eal-wo ld case s udy on he Ronda de Dal
in Ba celona.
•
Benchma k Con olle s: The s anda d ALINEA and Ups eam ALINEA
amp me e ing s a egies a e explained, along wi h de ails on senso placemen s and
algo i hmic logic.
•
P oposed Algo i hm: The op imized EqALINEA s a egy is in oduced, de ail-
ing how i imp o es conges ion managemen and ai ness in a ic low dis ibu ion.
•
E alua ion F amewo k: The me hodology o assessing ai ness and e iciency is
ou lined, including pe o mance me ics and compa a i e analysis be ween di e en
con ol s a egies.
3.1 Simula ion Models
This sec ion p esen s he wo simula ion en i onmen s used in his s udy:
•
A Toy Model: Se es as a con olled es en i onmen o e alua e amp me e ing
s a egies in a simpli ied se ing be o e applying hem o eal-wo ld condi ions.
•
A Real-Wo ld Case S udy: Focused on he Ronda de Dal , a majo u ban
eeway in Ba celona, whe e eal a ic condi ions a e simula ed.
The oy model allows o apid expe imen a ion and es ing o di e en con ol s a egies
33
Fai ness in Ramp Me e ing Feb ua y 2025
wi hou he complexi ies o eal-wo ld calib a ion. Once key insigh s a e ob ained, he
me hodologies a e applied o he mo e complex Ronda de Dal scena io o assess hei
e ec i eness in eal-wo ld a ic condi ions.
3.1.1 Toy Model
The oy model is a simpli ied a ic simula ion used o in es iga e he undamen al
dynamics o amp me e ing unde con olled condi ions. I p o ides a s uc u ed es
en i onmen o alida e he e iciency and ai ness o me e ing s a egies be o e deploying
hem in eal-wo ld ne wo ks.
The ne wo k consis s o a 4.1 km eeway segmen wi h a maximum speed o 100 km/h
(27.78 m/s) and h ee 500-me e -long on- amps a 1.0 km, 2.0 km, and 3.0 km om he
mainline’s s a ing poin .
Each on- amp is equipped wi h:
•A amp me e ing signal, loca ed 250 me e s be o e he me ge nose.
•Ademand loop and passage loop o measu e in lows.
•
A a ic senso 40 me e s downs eam
4
o each me ge poin o eco d mainline
occupancy.
Toy Model Ne wo k
To egula e amp me e ing, he ALINEA con ol s a egy is applied, adjus ing me e ing
a es based on eal- ime occupancy le els. The oy model’s ne wo k layou is shown in
Fig. 7.
Figu e 7: Toy Model Ne wo k.
4Adequa e dis ance a he Boule a d Pé iphé ique in Pa is Papageo giou e al. (1991).
34

Fai ness in Ramp Me e ing Feb ua y 2025
3.1.2 Ronda de Dal (Ba celona)
The Ronda de Dal is one o he mos c i ical oad in as uc u es in Ba celona. Unlike
adi ional highways ha p ima ily se e in e ci y commu e s, he Ronda de Dal plays a
mul i unc ional ole by managing u ban, me opoli an, and egional a ic wi hin a dense
u ban en i onmen .
Figu e 8: Les Rondes de Ba celona.
Cha ac e is ics o he Ronda / Ronda de Dal Along wi h he Ronda Li o al, he Ronda
de Dal o ms pa o Ba celona’s p ima y ing oad sys em, Les Rondes, which we e
cons uc ed o imp o e ci ywide mobili y by edis ibu ing ehicle lows away om he
u ban co e. Thei impac is pa icula ly e iden in he way hey ha e eo ganized a ic
dis ibu ion in and a ound Ba celona. The Rondes play eigh undamen al oles in he
ci y’s anspo a ion sys em (Ba celona Regional, 2019):
•
Les Rondes connec en: The Rondes acili a e mobili y be ween Ba celona and i s
su ounding municipali ies, se ing bo h occupa ional and pe sonal a el pu poses.
Acco ding o a ic da a, 69% o ips a e occupa ional (p ima ily wo k- ela ed),
while 31% a e pe sonal (including endez ous, shopping and leisu e). Addi ionally,
65% o ehicles ha e a single occupan , highligh ing he dominance o p i a e ca
35
Fai ness in Ramp Me e ing Feb ua y 2025
usage in daily mobili y
Figu e 9: T a el Pu pose & Vehicle Occupancy: Dis ibu ion on Ronda de Dal .
•
Les Rondes dis ibueixen: The Rondes dis ibu e a ic lows, educing con-
ges ion in he ci y cen e . O he 536,000 ehicles en e ing Ba celona daily, 60%
a e abso bed by he Rondes, easing p essu e on main u ban oads like G an Via,
Diagonal, Me idiana, and Via Augus a. Beyond egional a ic dis ibu ion, he
Rondes also s uc u e local mobili y, connec ing neighbo hoods ha ha e limi ed
public anspo accessibili y, pa icula ly o me opoli an and egional connec ions.
•
Les Rondes canali zen: The Rondes unc ion as high-capaci y u ban oads,
dis inc om con en ional highways. Thei design inco po a es equen en y and
exi poin s, s eepe g adien s, and a sinuous layou , p io i izing in eg a ion wi h he
exis ing u ban ab ic.
O iginally planned o espond o local mobili y needs, hei inal alignmen was
shaped by esiden ial demands, ensu ing accessibili y o su ounding neighbo hoods.
Howe e , hese design choices also in luence a ic in ensi y, speed, and conges ion
dynamics, making he Rondes a hyb id cha ac e is ics be ween highways and u ban
s ee s.
Figu e 10: T a ic Volume (Weekday).
36
Fai ness in Ramp Me e ing Feb ua y 2025
•
Les Rondes cohabi en: The Rondes in e sec densely popula ed u ban a eas,
in luencing he daily li es o app oxima ely 310,000 esiden s. The su ounding
neighbo hoods exhibi high socioeconomic con as s, anging om a eas wi h highe
pu chasing powe o some o Ba celona’s mos ulne able dis ic s, pa icula ly in
moun ainous zones.
Despi e hese dispa i ies, he Rondes o m pa o daily mobili y ou es** o
esiden s, connec ing hem o essen ial se ices such as supe ma ke s, heal hca e
cen e s, childca e acili ies, and communi y spaces.
•
Les Rondes con aminen: The Rondes con ibu e o ai pollu ion and noise
emissions, posing en i onmen al challenges ha need o be add essed h ough
sus ainable mobili y solu ions.
•
Les Rondes condicionen: The Rondes ac as a physical and social bounda y,
limi ing c oss-neighbo hood connec i i y and in luencing u ban de elopmen . They
de ine he ansi ion be ween he ci y and na u al spaces like Collse ola, Besòs, and
Llob ega , shaping land use and accessibili y.
•
Les Rondes e olucionen: The Rondes mus adap o e ol ing mobili y needs,
balancing p i a e ehicle use wi h sus ainable al e na i es. Public policies inc eas-
ingly p io i ize public anspo and non-mo o ized mobili y, aiming o educe ca
dependency and i s nega i e ex e nali ies.
•
Les Rondes enllacen: Ini ially designed as pe iphe al in as uc u e, he Ron-
des ha e ede ined Ba celona’s u ban bounda ies, linking he ci y cen e wi h i s
me opoli an su oundings. Today, hey connec majo economic hubs, heal hca e
acili ies, uni e si ies, and spo s a eas, in eg a ing Ba celona wi hin a b oade
egional amewo k.
Figu e 11: Economic Poles & U ban Cen ali ies.
37
Fai ness in Ramp Me e ing Feb ua y 2025
Why he Simula ion Focuses on Ronda de Dal (Di ec ion Llob ega ) This s udy speci i-
cally simula es he Llob ega -bound di ec ion o he Ronda de Dal . The selec ion was
in luenced by he ollowing ac o s:
•
Pe sonal Commu e Expe ience: O e he pas wo yea s, his sec ion has been
pa o he esea che ’s daily commu e, p o iding di ec insigh s in o i s a ic
beha io and conges ion pa e ns.
•
Pe sis en Conges ion and Inc easing Demand: The Llob ega -bound di ec ion
is one o he mos conges ed u ban highways in Ba celona, equen ly eaching
sa u a ion le els, especially du ing peak hou s. Despi e a 19% educ ion in a ic in
Ba celona’s cen al a enues, he Ronda de Dal has expe ienced a s eady inc ease
in demand. In he pas wo yea s alone, a ic along his co ido has g own by
10.62% (see Fig. 12), in ensi ying p essu e on exis ing in as uc u e.
•In as uc u e Limi a ions: Unlike o he me opoli an highways, he Ronda de
Dal was buil wi hin a cons ained u ban oo p in , wi h na ow lanes, equen on-
amps, and limi ed eme gency shoulde s. Physical expansion is un easible/un ealis ic,
as no ed by Ja ie O igosa, head o he Me opoli an U ban Planning O ice, who
s a es ha " he u ban ab ic se s he limi —expanding he Ronda would no
educe conges ion, as ehicles s ill need o en e es ic ed ci y a eas." Fu he mo e,
sus ainable mobili y policies, aimed a discou aging p i a e ehicle use in
Ba celona, ha e inad e en ly shi ed mo e conges ion o he Rondes, ein o cing
he need o demand-side solu ions like amp me e ing.
•
Policy and U ban Mobili y Objec i es: The Ba celona Ci y Council has ac i ely
pu sued policies o educe p i a e ehicle dependency in a o o sus ainable mobili y
op ions, such as public anspo and g een co ido s. Howe e , a signi ican po ion
o commu e s con inue o ely on ca s, leading o a pa adox whe e a ic es ic ions
in he ci y cen e push mo e ehicles on o he Rondes.
38
Fai ness in Ramp Me e ing Feb ua y 2025
–A2 – O - amp (1 lane).
–A3 – O - amp (1 lane).
–A4 – O - amp (1 lane).
–A5 – Mainline exi (3 lanes).
•Sequence : – E1 - A1 - A2- E2 - A3 - E3 - A4 - E4 - E4b - A5 - E5 -
Dynamic Flow Replica ion Based on T a ic In ensi y: To eplica e ealis ic a ic
a ia ions h oughou he mo ning peak pe iod, low spawning ollows he obse ed a ic
in ensi y pa e n (see Fig. 10) o e he cou se o a weekday.
•
The spawning a e is upda ed e e y 10 minu es o e lec eal- ime changes in a ic
demand.
•
The simula ion pe iod s a s a 06:00 AM and ends a 09:00 AM, co e ing he
mo ning conges ion buildup. A o al simula ion ime o 10.800 seconds.
•
Peak conges ion is expec ed o occu a ound 08:00 AM, in alignmen wi h eal-wo ld
a ic pa e ns.
3.2.3 Vehicle & D i e Popula ion
To simula e ealis ic d i ing beha io s, he simula ion u ilizes en ehicle ypes, di ided
in o i e passenge ca ca ego ies and i e mo o cycle ca ego ies. These ehicle ypes a e
cha ac e ized by a ying le els o d i ing agg essi eness, which in luence accele a ion,
decele a ion, lane-changing endencies, and o e all in e ac ion wi h su ounding a ic.
The simula ion now implemen s he SL2015 lane change model which in oduces a
mo e dynamic and ealis ic ep esen a ion o lane-changing beha io s based on d i e
agg essi eness, coope a ion, and an icipa ion.
Vehicle Type Classi ica ion:
•
Vehicles a e ca ego ized in o i e agg ession le els (agg 1 o agg 5), anging om
leas o mos agg essi e. Highe agg ession le els co espond o educed eac ion
imes, inc eased impa ience, as e speed adap a ion, and iskie lane-changing
beha io , including lane-spli ing o mo o cycles. The key cha ac e is ics o each
ehicle ype a e summa ized in Table 3.
45

Fai ness in Ramp Me e ing Feb ua y 2025
Each ehicle ype is de ined by speci ic beha io al pa ame e s, as de ailed below.
•Lane Change S a egy and Coope a ion
–
lcS a egic: De e mines how ea ly a ehicle plans a lane change. Highe alues
esul in ea lie decision-making o swi ch lanes s a egically.
–
lcCoope a i e: Con ols he willingness o yield o me ging ehicles. Lowe
alues indica e less coope a i e beha io , making lane changes mo e compe i i e.
–
lcSpeedGain: De ines how eage a ehicle is o change lanes o speed bene i s.
Highe alues lead o mo e equen lane changes o main ain op imal speed.
–
lcKeepRigh : De e mines he willingness o ollow keep- igh ules. Highe
alues esul in ea lie lane changes o he igh , while 0 disables his beha io .
–
lcPushy: De ines he willingness o o ce a lane change e en i gaps a e igh .
Highe alues indica e mo e agg essi e beha io .
•Sublane and La e al Beha io (Speci ic o SL2015)
–
lcSublane: Con ols he eage ness o use la e al posi ioning wi hin a lane.
Highe alues indica e mo e equen small la e al shi s, which is especially
ele an o mo o cycles.
–
minGapLa : Speci ies he minimum la e al gap be ween ehicles when using
sublane posi ioning.
–
lcPushyGap: De ines he minimum la e al gap equi ed o an agg essi e lane
change.
•Lane Change Timing and Impa ience
–
lcTimeToImpa ience: Speci ies he ime equi ed o a ehicle o each i s
maximum impa ience le el when i s lane change is blocked. A lowe alue leads
o quicke agg essi e beha io .
–
lcImpa ience: Adjus s lcPushy and lcAsse i e dynamically, meaning ha
ehicles become mo e agg essi e o e ime i hey canno change lanes.
•Accele a ion and Decele a ion Beha io
–
accel (m/s²): De ines he maximum accele a ion a e, a ec ing how quickly
ehicles each highe speeds.
–
decel (m/s²): De ines he maximum decele a ion (b aking) capaci y o a ehicle.
46
Fai ness in Ramp Me e ing Feb ua y 2025
•Maximum Speed and Speed Adap a ion
–
speedFac o : Mul iplie o he de aul speed limi . A alue o 1.0 means he
ehicle ollows he speed limi exac ly, while lowe alues indica e slowe d i ing
endencies.
–
speedDe : Adds a iabili y o ehicle speeds, ensu ing mo e ealis ic d i ing
beha io by simula ing di e en d i e endencies.
47
Fai ness in Ramp Me e ing Feb ua y 2025
Table 3: Upda ed Vehicle Type Cha ac e is ics
Pa am ca 1 ca 2 ca 3 ca 4 ca 5 mo 1 mo 2 mo 3 mo 4 mo 5
Class P P P P P M M M M M
speedFac o 0.80 0.85 0.90 0.95 1.00 0.85 0.90 0.93 0.95 1.00
speedDe 0.08 0.09 0.10 0.12 0.15 0.07 0.08 0.09 0.10 0.10
au 5.0 4.5 3.5 3.0 2.5 4.0 3.5 3.0 2.5 1.5
impa ience 2.0 2.5 3.0 3.5 4.0 2.0 2.8 3.8 4.5 5.0
lcImpa ience 1.0 1.5 2.0 2.5 3.0 1.2 1.8 2.4 3.0 3.6
lcCoope a i e 0.5 0.3 0.2 0.1 0.0 0.5 0.3 0.2 0.1 0.0
lcS a egic 60 40 25 15 5 50 35 20 10 5
lcSpeedGain 7.0 8.0 9.0 10.0 11.0 8.0 9.0 10.0 11.0 10.0
lcKeepRigh 0.6 0.5 0.3 0.2 0.0 0.6 0.4 0.3 0.15 0.0
lcPushy 1.2 1.8 2.2 2.6 3.0 1.5 2.0 2.5 3.0 3.5
lcPushyGap 0.5 0.4 0.3 0.2 0.1 0.5 0.4 0.3 0.2 0.1
lcSublane 6.0 5.0 4.0 3.0 2.0 5.0 4.0 3.0 2.0 1.0
lcAsse i e 2.0 2.8 3.5 4.2 5.0 2.5 3.0 3.8 4.5 6.0
lcTimeToImpa ience 15 12 9 6 3 15 12 9 6 3
lcSigma 0.4 0.6 0.8 1.0 1.2 0.5 0.7 0.9 1.1 1.3
minGapLa 0.8 0.6 0.5 0.4 0.3 0.6 0.5 0.4 0.3 0.2
maxSpeed 35 35 40 40 45 25 30 35 40 50
accel 1.8 2.2 2.5 2.8 3.0 3.5 4.0 4.5 5.0 6.0
decel 3.0 4.0 5.0 6.0 8.0 4.5 6.0 7.0 9.0 12.0
48
Fai ness in Ramp Me e ing Feb ua y 2025
3.3 Benchma k Con olle s
Benchma k con olle s a e implemen ed o p o ide a e e ence o e alua ing amp me e ing
s a egies. This s udy conside s ALINEA and Ups eam ALINEA, wo widely applied
amp me e ing echniques, o assess hei impac on a ic dynamics and compa e hei
pe o mance wi h he p oposed EqALINEA s a egy.
3.3.1 ALINEA and Ups eam ALINEA
Why ALINEA and Ups eam ALINEA as Benchma k Con olle s?
Since his hesis ocuses on ai ness in amp me e ing, he objec i e is o implemen
a widely used and e icien con ol s a egy as a baseline o compa ison. ALINEA was
selec ed because i is one o he mos commonly applied local amp me e ing algo i hms,
p o iding a solid ounda ion o he de elopmen and analysis o EqALINEA om a
ai ness pe spec i e.
Howe e , ALINEA aces wo c i ical limi a ions:
•Neglec o Fai ness Conside a ions:
ALINEA p io i izes mainline h oughpu bu does no explici ly conside equi y
among oad use s. Vehicles a ups eam on- amps, pa icula ly hose nea bo lenecks
o en endu e disp opo iona e delays, as he algo i hm allows un es ic ed in low
when pe cei ed downs eam condi ions appea unconges ed.
•Spa ial Sensing Cons ain s - No Ups eam Feedback:
ALINEA elies solely on downs eam senso s, making i incapable o de ec ing
conges ion p opaga ion ups eam o he me ge poin . Me ging con lic s a he amp
i sel may c ea e queues ex ending backwa d (ups eam), ye i he downs eam
occupancy emains below he h eshold, he sys em ails o igge es ic i e
me e ing. This eedback loop can in ensi y conges ion ins ead o mi iga ing i .
49
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 16: ALINEA Wo king P inciple (Scheme).
Sou ce: G egu ić e al. (2016)
Gi en hese challenges, Ups eam ALINEA was also selec ed as a benchma k because i
ex ends ALINEA’s logic by using ups eam occupancy measu emen s, allowing ea lie
conges ion de ec ion and be e delay dis ibu ion ac oss mul iple amps.
This decision is u he jus i ied by h ee key ac o s:
1.
Case S udy Con ex : The Ronda de Dal in Ba celona is a non- adi ional u ban
highway, cha ac e ized by equen on- amps, cons ained geome y, and mixed
a ic dynamics, as discussed in Sec ion 3.1.2. The applicabili y o adi ional
highway me e ing s a egies mus be econside ed in such an u ban se ing.
2.
Lack o a S anda dized De ec o Placemen o ALINEA: The e is no speci y
op imal senso loca ions, ALINEA has been applied wi h a ied senso placemen s
in di e en case s udies:
•
Luaibi e al. (2023) ound ha 300m downs eam om he me ge nose p o ided
an op imal c i ical occupancy de ec ion poin .
•
Papageo giou e al. (1991) placed senso s 40m downs eam a he Boule a d
Pé iphé ique in Pa is and 400m downs eam in Ams e dam.
These a ia ions indica e ha ALINEA’s senso placemen is no uni e sally de ined
and choosing he op imum posi ion o he downs eam de ec o s’ s a ion is di icul .
3.
P eceden om O he Con ol S a egies: Many o he s amp me e ing ap-
p oaches inco po a e ups eam occupancy o enhance con ol e ec i eness. Table 14
summa izes exis ing s a egies ha u ilize ups eam in o ma ion. Following his
50

Fai ness in Ramp Me e ing Feb ua y 2025
app oach, implemen ing Ups eam ALINEA ep esen s a na u al ex ension in he
de elopmen o ALINEA-based s a egies.
These con olle s se e wo p ima y objec i es:
1.
Es ablishing a Pe o mance Baseline - The e alua ion o ALINEA and Ups eam
ALINEA p o ides a e e ence o assessing he e ec i eness o EqALINEA. By
simula ing hese con olle s unde iden ical a ic condi ions, hei capaci y o
educe conges ion and con ol eeway access can be sys ema ically compa ed.
2.
Assessing T a ic E iciency and Fai ness - Local ALINEA con ol s a egy is
designed o maximize mainline e iciency, bu i does no explici ly conside ai ness
among oad use s, pa icula ly on- amp ehicles, which may expe ience excessi e
delays specially o hose use s nea o he bo e neck.
Ups eam ALINEA ex ends ALINEA’s logic by inco po a ing ups eam occupancy
o p e en conges ion om p opaga ing; howe e , i does no ully add ess ai ness
conce ns. E alua ing hese con olle s allows o an analysis o :
•Thei e ec i eness in managing eeway capaci y o h oughpu .
•The impac on on- amp ehicle delays.
•
The ex en o which ai ness imp o emen s could be achie ed wi hou comp o-
mising e iciency.
The ollowing sec ions desc ibe he senso con igu a ions, con ol logic, and ope a ional
p inciples o ALINEA and Ups eam ALINEA, which se e as he ounda ion o de eloping
a amp me e ing app oach wi h a ocus on ai ness.
3.3.2 Componen Placemen and Da a Collec ion
Types o De ec o s Used
The simula ion employs LaneA ea (e2) de ec o s om SUMO o collec eal- ime
a ic da a o amp me e ing ope a ions. P o iding key a ic pa ame e s, including
occupancy, speed, queue leng h, and me e ing low pe signal cycle.
•Occupancy: The pe cen age o he de ec o ’s leng h ha is occupied by ehicles.
•Speed: The a e age eloci y o de ec ed ehicles o e de ec o s.
51
Fai ness in Ramp Me e ing Feb ua y 2025
•Queue Leng h: The leng h o ehicles wai ing on amps.
•Me e ed low: measu ing he numbe o ehicles ha has le on- amp.
Signage, De ec o ’s Loca ion, and Func ions
Ramp Me e ing Signage - Ramp me e ing in he simula ion is con olled using a ic
signals ha egula e ehicle en y on o he eeway. The signals ope a e based on con ol
s a egy algo i hms, dynamically adjus ing g een ime du a ions o con ol he numbe o
ehicles me ging on o he mainline. The me e ing a e, which de e mines he sha e o
g een ime wi hin each con ol cycle (e.g., e e y 30 seconds o 1 minu e), is ecalcula ed a
he end o each in e al based on eal- ime a ic condi ions, ensu ing adap i e egula ion
o amp in lows.
•
T a ic Ligh Placemen o Toy Model: Posi ioned 250 me e s ups eam o
he me ge nose on each on- amp.
•
T a ic Ligh Placemen o Ronda de Dal : Posi ioned 30 me e s ups eam
o he me ge nose on each on- amp.
The de ec o s a e s a egically placed o cap u e ele an a ic condi ions. The main
ca ego ies o senso s and hei unc ions a e as ollows:
•Up-Ramp De ec o :
–
Placed be o e he amp me e ing signal o moni o he o al leng h o wai ing
ehicles and assess on- amp demand.
–
P o ides inpu o de e mining he equi ed me e ing a e adjus men s based
on conges ion le els o EqALINEA con ol s a egy.
•Down-Ramp De ec o :
–Placed a e he Ramp me e ing signal, acking he me ging low a es.
•Downs eam F eeway De ec o s:
–Placed 40 me e s a e he me ge poin (o "me ge nose") o moni o eeway
conges ion.
–
These de ec o s supply occupancy da a o ALINEA, which egula es amp
me e ing based on eeway condi ions.
•Ups eam F eeway De ec o s:
–
Posi ioned a hund ed me e s be o e he amps (speci ic o Ups eam ALINEA).
–
Used o an icipa e conges ion o ma ion and egula e amp me e ing acco dingly,
p e en ing conges ion om p opaga ing.
52
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 17: Ramp Me e ing Componen Placemen .
Sou ce: Own Souce
3.3.3 Implemen a ion (ALINEA & Ups eam ALINEA)
The ALINEA algo i hm is implemen ed h ough he alinea_con ol unc ion, which
adjus s he amp me e ing signal imings based on eal- ime a ic condi ions. Below is
an explana ion o how he o mula is applied in he code:
53
Fai ness in Ramp Me e ing Feb ua y 2025
ALINEA Con ol Logic The undamen al con ol equa ion:
(k) = (k−1) + KR[ˆo−oou (k)] (1)
Whe e:
• (k): Me e ing a e a he cu en ime cycle.
• (k-1): Me e ing a e om he p e ious ime cycle.
•KR
: Con ol gain pa ame e ( eh/h). In eal-li e expe imen s, a alue o 70 eh/h
was ound e ec i e. Fo simula ion wi h a 1-minu e cycle pe iod, i is no malized
as
KR
= 70/60 eh/min. I in luences how agg essi ely he algo i hm adjus s he
me e ing a e in esponse o changes in a ic condi ions.
•ˆo: Ta ge occupancy (se o 20% in his s udy).
•oou
(
k
): Measu ed downs eam occupancy (Exi _O_Measu ed) a he cu en ime
s ep.
Implemen a ion De ails wi h T aci
1de alinea_con ol(in e sec ion, Q_p e ious_ a e, O_Ta ge , Exi _O_Measu ed,
g een_sha es):,→
2"""
3Alinea amp me e ing con ol logic.
4
5Pa ame e s:
6- in e sec ion: Ramp me e ing loca ion ID
7- Q_p e ious_ a e: P e ious low a e o me e ing con ol
8- O_Ta ge : Ta ge eeway occupancy
9- Exi _O_Measu ed: Measu ed downs eam occupancy
10 - g een_sha es: P e ious g een ime alloca ion lis
11
12 Re u ns:
13 - Q_ a e: Upda ed me e ing a e
14 - GREEN_SHARE: Adjus ed g een ime alloca ion
15 """
16 Q_ a e =Q_p e ious_ a e +K *(O_Ta ge -Exi _O_Measu ed)
17
54
Fai ness in Ramp Me e ing Feb ua y 2025
ions, quan i ying ip comple ion e iciency.
•
To al T a el Time (s): Sum o ime aken by ehicles o each des ina ions,
wi h lowe alues indica ing e icien ou ing.
•
To al T a el Dis ance (km): To al dis ance a eled by ehicles, whe e sho e
dis ances sugges minimal de ou s due o conges ion.
•
A e age Speed o Vehicles(km/h): Mean speed o ehicles, e lec ing o e all
a ic luidi y o comple ed ips.
•
To al Delays (s): Sum o addi ional ime spen due o conges ion o me e ing,
di ec ly quan i ying conges ion’s impac .
•
A e age Delay pe Vehicle (min/ eh): De ia ion be ween ac ual a el ime
and ee- low a el ime, measu ing he ime expe ienced pe ehicle due o
conges ion based on di e en scena io.
•
Mainline Occupancy (%): The p opo ion o oad space occupied by ehicles
a a gi en ime, used o measu e conges ion le els. High occupancy indica es
conges ion; op imal occupancy ensu es smoo h low (e.g., 10–20% is ideal o
ee- low condi ions).
•
A e age F ee-Flow Di e ence (s): De ia ion be ween ac ual a el ime and
ee- low a el ime.
2.
Fai ness Conside a ions: How equi ably he me e ing delays a e dis ibu ed
among on- amp use s. As men ioned in Sec ion 3.4.1, in his s udy we combine
Rawlsian ai ness wi h egali a ian ai ness; ou aim is o minimize a iabili y in he
delay me ics ac oss amps, which ansla es in o a lowe Gini coe icien .
•
Gini Coe icien : Inequali y measu e (0–1) o delays among comple ed ips,
wi h 0 = pe ec ai ness. (Egali a ian Fai ness)
•
A e age Wai ing Time pe On-Ramp (s): Mean delay expe ienced by ehicles
a each on- amp, iden i ying bo lenecks. (Ha sanyian Fai ness)
•
Me ging Ra e pe On-Ramp ( eh/cycle): Vehicles en e ing he eeway pe
cycle om each on- amp, ensu ing no amp is disp opo iona ely es ic ed.
•
Numbe o Vehicles Joined o Highway ( eh): To al ehicles en e ing ia each
on- amp, ensu ing demand is me wi hou bias.
61

Fai ness in Ramp Me e ing Feb ua y 2025
•
Max On-Ramp Wai ing Time (s): Longes delay expe ienced a any single
on- amp, highligh ing wo s -case inequi ies. (Rawlsian Fai ness)
•
Red ime assigna ion pe On-Ramp (s/%): Du a ion/pe cen age o ed signals
pe on- amp, balancing me e ing es ic ions equi ably.
62
Fai ness in Ramp Me e ing Feb ua y 2025
4 Resul s and Discussion
This sec ion p esen s he simula ion esul s and e alua es he impac o ALINEA, Ups eam
ALINEA, and EqALINEA on a ic e iciency and ai ness. The analysis is based on wo
simula ion models, p e iously in oduced in Sec ion 3.2, which ep esen di e en le els o
ne wo k complexi y and a ic demand.
To ensu e cla i y, he esul s a e p esen ed in a laye ed app oach:
•Gene al ends a e i s analyzed using he Toy Model.
•
The key di e ences obse ed in he mo e complex Ronda de Dal Model a e hen
highligh ed.
•Compa a i e ables and isual o e lays complemen edundan da a p esen a ions.
4.1 T a ic E iciency Analysis
The e ec i eness o he benchma k con ol s a egies in managing eeway conges ion is
e alua ed using he a ic e iciency me ics ou lined in Sec ion 3.5. The pe o mance
o ALINEA and Ups eam ALINEA is hen compa ed agains he p oposed EqALINEA
algo i hm o assess imp o emen s in a ic low, delay educ ion, and o e all sys em
e iciency.
The analysis is conduc ed o e di e en simula ion du a ions (ST = Simula ion Time)
o each model. The ini ial phase is conside ed as ansi o y esponde pe iod, whe e
he sys em g adually g adually ills wi h ehicles ans s abilizes. To ensu e he accu a e
e alua ion, only he s eady-s a e po ion o he simula ion is analyzed:
•Toy Model: ST = 1000s o 3000s
•Ronda Toy Model: ST = 1000s o 10800s
The a ic e iciency me ics e eal dis inc pe o mance ends ac oss con ol s a egies
in bo h he Toy and Ronda models.
63
Fai ness in Ramp Me e ing Feb ua y 2025
Table 4: T a ic E iciency Me ics o Toy Model
Me ic NoCon ol ALINEA UpALINEA EqALINEA
TDV ( eh/ST) 1795 1657 1601 2097
TAV ( eh/ST) 1212 1211 1209 1219
AR (%) 67.52 73.08 75.52 58.13
TTT (h) 294 256 223 277
TTD (km) 4636 4714 4891 4910
AS (km/h) 15.7 18.4 21.9 17.7
TD (h) 240 205 171 210
AD (min/ eh) 8.02 7.42 6.41 6.01
Table 5: T a ic E iciency Me ics o Ronda Model
Me ic NoCon ol ALINEA UpALINEA EqALINEA
TDV ( eh/ST) 11090 10731 10558 10402
TAV ( eh/ST) 10285 10125 9989 9942
AR (%) 92.74 94.35 94.61 95.58
TTT (h) 2182 1761 1648 1454
TTD (km) 36654 39923 40932 41150
AS (km/h) 16.8 22.7 24.8 28.3
TD (h) 1706 1247 1123 927
AD (min/ eh) 9.23 6.97 6.38 5.35
In he Toy Model, ALINEA and Ups eam ALINEA (UpALINEA) exhibi measu able
imp o emen s o e he NoCon ol baseline. ALINEA educes o al delays (TD) by 14.6%
(240 h o 205 h) and inc eases he a i al a e (AR) by 5.6 pe cen age poin s (67.5% o
73.1%). UpALINEA u he enhances hese ou comes, lowe ing delays by 28.8% (240 h
o 171 h) and achie ing he highes AR (75.5%). Howe e , EqALINEA demons a es
a con as ing end: i s AR declines o 58.1%, and o al a el ime (TTT) inc eases
ma ginally (277 h s. NoCon ol’s 294 h). This di e gence a ises om EqALINEA’s
p io i iza ion o equi able me ging, which educes pe - ehicle delays (AD: 6.01 min/ eh s.
NoCon ol’s 8.02 min/ eh) bu inc eases queuing a speci ic on- amps (e.g., J11), leading
o highe sys em-wide ehicle accumula ion.
In he Ronda Model, all s a egies ou pe o m NoCon ol, wi h EqALINEA deli e ing
he mos balanced esul s. EqALINEA educes o al delays by 45.7% (1,706 h o 927 h),
achie es he highes a e age speed (28.3 km/h s. NoCon ol’s 16.8 km/h), and main ains
he highes AR (95.6%). UpALINEA also pe o ms obus ly, educing delays by 34.1%
(1,706 h o 1,123 h) and imp o ing a e age speed o 24.8 km/h. ALINEA, while e ec i e,
lags behind hese wo s a egies, unde sco ing i s limi a ions in managing complex ne wo k
64
Fai ness in Ramp Me e ing Feb ua y 2025
in e ac ions.
Bo h models e eal inc eased o al a el dis ances (TTD) unde con olled s a egies,
which can be explained by he in luence o amp me e ing con ol. Howe e , hese inc eases
a e coun e balanced by signi ican gains in speed and delay educ ion.
Fo mo e de ail in o ma ion e e o Table 4 o Toy Model me ics and Table 5 o Ronda
de Dal Model esul s.
4.2 Fai ness E alua ion
4.2.1 Gini Coe icien Analysis
The Gini coe icien is used o assess inequali y in delay dis ibu ion, whe e lowe alues
indica e g ea e ai ness. Table 6 p esen s he Gini coe icien s o a i ed ehicles, depa ed
ehicles, and ehicles s ill in ansi unde each amp me e ing s a egy o bo h he Toy
Model and Ronda De Dal Model.
Table 6: Gini Coe icien s o Toy Model and Ronda De Dal Model
Toy Model
Con ol Gini A i ed Gini Depa ed Gini on Sys em
NoCon ol 0.44 0.43 0.34
ALINEA 0.43 0.43 0.38
UpS eamALINEA 0.35 0.37 0.40
EqALINEA 0.41 0.40 0.35
Ronda Model
Con ol Gini A i ed Gini Depa ed Gini on Sys em
NoCon ol 0.40 0.40 0.39
ALINEA 0.29 0.30 0.36
UpS eamALINEA 0.25 0.26 0.34
EqALINEA 0.21 0.22 0.31
65
Fai ness in Ramp Me e ing Feb ua y 2025
Fai ness T ends in he Toy Model
In he Toy Model, ALINEA exhibi s minimal ai ness imp o emen s compa ed o NoCon-
ol, as he Gini coe icien o a i ed ehicles emains nea ly unchanged (0.43 s. 0.44 in
NoCon ol). Howe e , i s on-sys em Gini coe icien inc eases (0.38 s. NoCon ol’s 0.34),
indica ing ha inequali y in delay accumula ion wo sens o ehicles s ill in ansi . This
is due o ALINEA’s ocus on maximizing eeway h oughpu , which sac i ices ai ness
o on- amp use s downs eam o bo lenecks (J9 and J10), whe e conges ion p opaga es
apidly. As shown in Fig. 19 and Fig. 20, he ALINEA con ol s a egy comple ely s ops
me e ing o amps J9 and J10 om simula ion ime 1700s onwa d, p io i izing mainline
low bu disp opo iona ely delaying on- amp use s a hese loca ions.
Ups eam ALINEA achie es a mo e balanced delay dis ibu ion, educing he Gini
coe icien o a i ed ehicles o 0.35. This imp o emen occu s because me e ing is
ac i a ed ea lie , dis ibu ing delays mo e e enly o e ime as conges ion o ms and
p opaga es. Howe e , i s on-sys em Gini coe icien emains high (0.40), sugges ing ha
while ai ness imp o es o comple ed ips, delay a iabili y pe sis s o ehicles s ill in
ansi . As depic ed in Fig. 19 and Fig. 20, he Ups eam ALINEA con ol s a egy s a s
assigning mo e han 50% ed ime o amps J9 and J10 as ea ly as simula ion ime 1250s,
e ec i ely delaying conges ion p opaga ion ups eam. As a esul , e en hough Ups eam
ALINEA ollows he same con ol logic as ALINEA, i a oids comple ely blocking access
o amps J9 and J10, ensu ing mo e equi able on- amp access while s ill managing eeway
conges ion.
EqALINEA success ully balances e iciency and ai ness, achie ing he lowes on-sys em
Gini coe icien (0.35), ensu ing ha wai ing imes a e mo e equi ably dis ibu ed ac oss
all amps. Howe e , he Gini coe icien o a i ed (0.41) and depa ed ehicles (0.40)
emains mode a e, e lec ing a delibe a e ade-o be ween main aining eeway e iciency
and ensu ing equi able amp access. As seen in Fig. 19 and Fig. 20, EqALINEA is he
only con ol s a egy ha con inues assigning me e ing a es o on- amp use s while
main aining sys em e iciency, e ec i ely mi iga ing ex eme delays wi hou comple ely
es ic ing access.
Fai ness T ends in he Ronda De Dal Model
The Ronda De Dal Model, wi h highe demand and a longe simula ion pe iod, shows
s onge ai ness imp o emen s ac oss all con ol s a egies. ALINEA educes delay in-
equali y compa ed o NoCon ol, bu on-sys em ai ness emains a challenge, as conges ion
66

Fai ness in Ramp Me e ing Feb ua y 2025
pe sis s o ehicles s ill in ansi .
Ups eam ALINEA u he imp o es ai ness, educing he Gini coe icien o 0.25 o
a i ed ehicles, as amp me e ing is ac i a ed ea lie , p e en ing ex eme bo lenecks
a indi idual amps. Howe e , EqALINEA deli e s he mos equi able esul s, wi h
Gini alues o 0.21 o a i ed ehicles and 0.22 o depa ed ehicles, demons a ing i s
e ec i eness in dis ibu ing delays ai ly ac oss all use s.
A no able end in he Ronda De Dal Model is ha on-sys em Gini coe icien s emain
high ac oss all s a egies (0.31–0.39). This occu s because he simula ion ex ends beyond
peak hou s (1000s o 10800s), allowing long-dis ance a ele s o expe ience educed
delays, which in luences he ai ness assessmen . Fu u e e alua ions could bene i om
dynamically analyzing on-sys em ai ness ela i e o a ic in ensi y o be e cap u e
ime-dependen equi y a ia ions.
4.2.2 On-Ramp Use Expe ience Analysis
The e alua ion o amp me e ing s a egies ocuses on hei abili y o balance e iciency and
equi y ac oss on- amps. Key me ics—maximum wai ing imes, me ging a es, ed ime
dis ibu ion, and success ul eeway en ies—highligh he ade-o s be ween benchma k
con ol me hods (ALINEA, UpALINEA) and he ai ness-o ien ed EqALINEA algo i hm.
In he ollowing analysis, key me ics will be examined o illus a e he he impac o
di e en amp me e ing s a egies on use expe ience a indi idual amps.
•Wai ing Time Dis ibu ion:
Bo h a e age and maximum wai ing imes (Tables 7 and 8) e eal EqALINEA’s
abili y o mi iga e ex eme delays while balancing a e age conges ion. In he Toy
Model, EqALINEA educes a e age wai ing imes by 53% o J9 (210.23 s s.
ALINEA’s 455.73 s) and 47% o J10 (216.63 s s. ALINEA’s 411.77 s), despi e
J11’s highe maximum wai (183 s s. ALINEA’s 33 s). This ade-o e lec s
a delibe a e edis ibu ion: J11’s maximum wai emains 44% lowe han unde
UpALINEA (380 s), while i s a e age wai dec eases by 39% (194.11 s s. 322.99 s),
demons a ing p io i ized ai ness o e localized op imiza ion.
67
Fai ness in Ramp Me e ing Feb ua y 2025
In he Ronda Model, EqALINEA na ows dispa i ies be ween amps. While E3 and
E5 expe ience inc eased a e age wai s (203.12 s and 186.82 s) compa ed o ALINEA
(105.70 s and 103.78 s), hei maximum wai s emain con olled (281 s and 250 s
s. ALINEA’s 95 s and 88 s). Con e sely, high-conges ion amps like E2 and E4b
see educed a e age wai s (213.52 s and 175.98 s) unde EqALINEA compa ed o
ALINEA (223.84 s and 188.16 s)bu E4 aces a inc emen om 175.83 s o 191.12 and
he maximum wai s o E2 and E4 dec easing by 56% and 50%, espec i ely. This
equilib ium p e en s sys emic bo lenecks, as no amp exceeds 280 s in maximum
delay -al hough, in heo y, he delay should no su pass 240s (4 minu es), which
equi es u he in es iga ion- (Table 8).
Table 7: A e age Wai ing Time pe On-Ramp [s]
Toy Model
Me ic NoCon ol ALINEA UpALINEA EqALINEA
J9 122.96 455.73 469.71 210.23
J10 152.34 411.77 482.18 216.63
J11 93.64 101.46 270.30 155.46
A e age 122.98 322.99 407.4 194.11
Ronda De Dal Model
Me ic NoCon ol ALINEA UpALINEA EqALINEA
E2 140.41 223.84 257.91 213.52
E3 71.77 105.70 147.25 203.12
E4 108.49 175.83 222.09 191.12
E4b 108.88 188.16 148.66 175.98
E5 60.85 103.78 149.55 186.82
A e age 98.08 159.46 185.09 194.11
68
Fai ness in Ramp Me e ing Feb ua y 2025
Table 8: Maximum Wai ing Time pe On-Ramp [s]
Toy Model
Ramp NoCon ol ALINEA UpALINEA EqALINEA
J9 400 1312 1337 308
J10 314 1224 1147 252
J11 12 33 380 183
Max 400 1312 1337 308
Ronda De Dal Model
Ramp NoCon ol ALINEA UpALINEA EqALINEA
E2 195 620 607 274
E3 20 95 202 281
E4 158 524 409 261
E4b 82 267 377 264
E5 14 88 183 250
Max 195 620 607 281
69
Fai ness in Ramp Me e ing Feb ua y 2025
•Me ging Ra e pe On-Ramp [ eh/cycle]:
While NoCon ol achie es he highes me ging a es (e.g., 10.86 eh/cycle a J9 in
he Toy Model; Table 9), i lacks egula ion, leading o uns ablesa u a ion a ic low.
ALINEA and UpALINEA supp ess me ging a es signi ican ly (e.g., 2.7 eh/cycle
a J9 o ALINEA), p io i izing eeway h oughpu o e amp accessibili y. EqA-
LINEA s ikes a middle g ound, imp o ing me ging a es compa ed o UpALINEA
(5.19 eh/cycle s. 2.58 eh/cycle a J9) while ensu ing no single amp domina es
access. This balance is pa icula ly e iden in he Ronda Model, whe e EqALINEA
na ows he gap be ween amps (3.36–4.07 eh/cycle) compa ed o ALINEA’s wide
dispa i y (3.15–5.64 eh/cycle).
Table 9: Me ging Ra e pe On-Ramp [ eh/cycle]
Toy Model
Ramp NoCon ol ALINEA UpALINEA EqALINEA
J9 10.44 2.7 2.58 5.19
J10 10.14 2.82 2.43 5.13
J11 10.86 7.8 4.23 6.39
A e age 10.48 4.44 3.08 5.57
Ronda de Dal Model
Ramp NoCon ol ALINEA UpALINEA EqALINEA
E2 8.03 3.78 3.37 3.90
E3 8.87 5.52 4.45 3.47
E4 7.40 3.15 3.86 3.36
E4b 8.52 4.34 3.61 4.07
E5 9.43 5.64 4.44 3.77
A e age 8.45 4.49 3.95 3.71
•Numbe o Vehicles Joined o Highway ( eh):
The numbe o ehicles success ully me ging on o he highway e lec s he balance
be ween conges ion con ol and equi able access (Table 10). In he Toy Model,
EqALINEA allows a mo e consis en en y a e ac oss amps (197–236 eh/ST), sig-
ni ican ly imp o ing upon UpALINEA (115–177 eh/ST) and ALINEA (130–274 e-
h/ST), which es ic access a ce ain amps. While ALINEA achie es sligh ly
highe h oughpu a J11, i does so a he expense o J9 and J10, which expe ience
subs an ial educ ions in me ging a es. EqALINEA, in con as , dis ibu es access
70
Fai ness in Ramp Me e ing Feb ua y 2025
o au onomous ehicle eadiness, pape p esen ed a he P oceedings o he 25 h ITS
Wo ld Cong ess, Copenhagen, Denma k, 17–21.
Kes en, A. S., M. E gün and T. Yai (2013) An analysis on e iciency and equi y o
ixed- ime amp me e ing, Jou nal o T anspo a ion Technologies,3(2A).
Ko sialos, A. and M. Papageo giou (2001) E iciency e sus ai ness in ne wo k-wide amp
me e ing, pape p esen ed a he ITSC 2001. 2001 IEEE In elligen T anspo a ion
Sys ems. P oceedings (Ca . No. 01TH8585), 1189–1194.
Ko sialos, A., M. Papageo giou, M. Mangeas and H. Haj-Salem (2002) Coo dina ed and
in eg a ed con ol o mo o way ne wo ks ia non-linea op imal con ol, T anspo a ion
Resea ch Pa C: Eme ging Technologies,10 (1) 65–84.
Långs öm, S. and E. F idsäll (2019) Op imizing a ic low on conges ed oads.
Le inson, D. and L. Zhang (2006) Ramp me e s on ial: E idence om he win ci ies
me e ing holiday, T anspo a ion Resea ch Pa A: Policy and P ac ice,40, 810–828,
02 2006.
Le inson, D., L. Zhang, S. Das and A. Sheikh (2002) Ramp me e s on ial: e idence
om he win ci ies amp me e s shu -o , pape p esen ed a he 81s T anspo a ion
Resea ch Boa d Con e ence, Washing on DC.
Le inson, D. M., L. Zhang and A. Sheikh (2004) Measu ing he equi y and e iciency o
amp me e s, Mas e Thesis, Uni e si y o Minneso a.
Li, D., P. Ranji ka and Y. Zhao (2016) E iciency and equi y pe o mance o a coo dina ed
amp me e ing algo i hm, P ome -T a ic&T anspo a ion,28 (5) 507–515.
Lipp, L. E., L. J. Co co an and G. A. Hickman (1991) Bene i s o cen al compu e con ol
o Den e amp-me e ing sys em, no. 1320 in 1320, T anspo a ion Resea ch Reco d,
0361-1981.
Liu, J.-C. S., J. L. Kim, Y. Chen, Y. Hao, S. Lee, T. Kim and M. Thomadakis (1994) An
Ad anced eal- ime amp me e ing sys em (ARMS): he sys em concep , ol. 1232 (24),
Texas T anspo a ion Ins i u e.
Lu, C., J. Huang, L. Deng and J. Gong (2017) Coo dina ed amp me e ing wi h equi y
77

Fai ness in Ramp Me e ing Feb ua y 2025
conside a ion using ein o cemen lea ning, Jou nal o T anspo a ion Enginee ing, Pa
A: Sys ems,143 (7) 04017028.
Luaibi, W. K., L. V. Leong and H. A. Al-Jameel (2023) Assessmen o alinea me hod
pe o mance a di e en loop de ec o loca ions using ield da a and mic o-simula ion
modeling ia aimsun, Open Enginee ing,13 (1) 20220475.
Meng, Q. and H. L. Khoo (2010) A pa e o-op imiza ion app oach o a ai amp me e ing,
T anspo a ion Resea ch Pa C: Eme ging Technologies,18 (4) 489–506.
Mizu a, A., K. Robe s, L. Jacobsen and N. Thompson (2014a) Ramp me e ing: A
p o en, cos -e ec i e ope a ional s a egy—a p ime , Technical Repo ,FHWA-HOP-
14-020, U.S. Depa men o T anspo a ion, Fede al Highway Adminis a ion, O ice
o Ope a ions, 1200 New Je sey A e S.E., Washing on, D.C. 20590, Oc obe 2014.
Accessed: 2024-10-16.
Mizu a, A., K. Robe s, L. Jacobsen, N. Thompson and J. Colya (2014b) Ramp me e ing:
a p o en, cos -e ec i e ope a ional s a egy: a p ime , Technical Repo , Uni ed S a es.
Fede al Highway Adminis a ion.
Municipali y, I. M. (n.d.) Pa icipa ion con ol applica ion,
h ps://web.a chi e.o g/
web/20170618130630/h p:/ km.ibb.go . /ku umsal/habe le - e-duyu ula /
ka ilim-kon olu-uygulamasi. Accessed: 2025-02-28.
Paesani, G., J. Ke , P. Pe o ich and F. Khos a i (1997) Sys em wide adap i e amp me-
e ing (swa m), pape p esen ed a he Me ging he T anspo a ion and Communica ions
Re olu ions. Abs ac s o ITS Ame ica Se en h Annual Mee ing and Exposi ionITS
Ame ica.
Papageo giou, M., J.-M. Blosse ille and H. Hadj-Salem (1990) Modelling and eal- ime
con ol o a ic low on he sou he n pa o boule a d pe iphe ique in pa is: Pa i:
Modelling, T anspo a ion Resea ch Pa A: Gene al,24 (5) 345–359.
Papageo giou, M., H. Hadj-Salem, J.-M. Blosse ille e al. (1991) Alinea: A local eedback
con ol law o on- amp me e ing, T anspo a ion esea ch eco d,1320 (1) 58–67.
Papageo giou, M. and A. Ko sialos (2002) F eeway amp me e ing: An o e iew, IEEE
ansac ions on in elligen anspo a ion sys ems,3(4) 271–281.
Papamichail, I., M. Papageo giou, V. Vong and J. Ga ney (2010) Heu is ic amp-me e ing
78
Fai ness in Ramp Me e ing Feb ua y 2025
coo dina ion s a egy implemen ed a monash eeway, aus alia, T anspo a ion Re-
sea ch Reco d,2178 (1) 10–20.
Riehl, K., A. Kou elas and M. Mak idis (2024) Towa ds ai oads–why we should & how
o imp o e he ai ness in a ic enginee ing, a Xi p ep in a Xi :2408.01309.
Shaaban, K., M. A. Khan and R. Hamila (2016) Li e a u e e iew o ad ancemen s in
adap i e amp me e ing, P ocedia Compu e Science,83, 203–211.
Sma agdis, E., M. Papageo giou and E. Kosma opoulos (2004) A low-maximizing adap i e
local amp me e ing s a egy, T anspo a ion Resea ch Pa B: Me hodological,38 (3)
251–270.
S ephanedes, Y. J. (1994) Implemen a ion o on-line zone con ol s a egies o op imal
amp me e ing in he minneapolis ing oad, Se en h In e na ional Con e ence on ‘Road
T a ic Moni o ing and Con ol.
Swi ze land, A. (2022) S a is ics o a ic de elopmen and a ic low 2021,
h ps://www.
as a.admin.ch/dam/as a/de/dokumen e/ab eilung_s assenne zeallgemein/
e keh sen wicklung-und- e uegba kei -de -na ionals assen-jah esbe ich -2022.
pd .download.pd /Ve keh sen wicklung_und_Ve %C3%BCgba kei _de _
Na ionals assen_Jah esbe ich _2022.pd . Accessed: 2025-02-28.
Taale, H., W. Schou en and J. an Koo e (1994) Design o a coo dina ed amp-me e ing
sys em nea ams e dam, pape p esen ed a he Se en h In e na ional Con e ence on
Road T a ic Moni o ing and Con ol, 185–189.
Tan, H., X. Yang and A. Wu (2013) The op imiza ion o on- amp junc ion capaci y based
on zippe ed con ol s a egy, P ocedia-Social and Beha io al Sciences,96, 1764–1773.
Taylo , C., D. Meld um and L. Jacobson (1998) Fuzzy amp me e ing: Design o e iew
and simula ion esul s, T anspo a ion Resea ch Reco d,1634 (1) 10–18.
ThePape (2022) T a ic signal con ol o amps,
h ps://m. hepape .cn/baijiahao_
19266072. Accessed: 2025-02-28.
T a ik e ke (2019) Ramp me e ing inal epo ,
h ps:// udin o. a ik e ke .
se/ udin oex e nwebb/Publika ione /Publika ione _003701_003800/
Publika ion_003764/190117_Ramps y ning%20slu appo _190117.pd
. Accessed:
2025-02-28.
79
Fai ness in Ramp Me e ing Feb ua y 2025
T anspo , A. (n.d.) Ramp signals,
h ps://a .go .nz/abou -us/
s ee -main enance/ amp-signals. Accessed: 2025-02-28.
T ubia, S., S. Cu o, S. Ba be i, A. Se e ino, F. A ena and G. Pau (2021) Analysis
and e alua ion o amp me e ing: F om his o ical e olu ion o he applica ion o new
algo i hms and enginee ing p inciples, Sus ainabili y,13 (2) 850.
Uni e si ä Duisbu g-Essen, d. (n.d.) Chap e 7,
h ps://duepublico2.uni-due.de/
se le s/MCRFileNodeSe le /duepublico_de i a e_00005368/08chap e 7.pd
.
Accessed: 2025-02-28.
an Lindonk, W. (2020) Speed limi s and hei e ec on eeway capaci y, Ph.D. Thesis,
MSc Thesis), Del Uni e si y o Technology, S e inweg 1, Del .
Wang, Y., Y. Kan, M. Papageo giou and I. Papamichail (2014) Local amp me e ing wi h
dis an downs eam bo lenecks: A compa a i e s udy, pape p esen ed a he 17 h
In e na ional IEEE Con e ence on In elligen T anspo a ion Sys ems (ITSC), 768–773.
Wang, Y., K. A. Pe ine and Y. Lao (2008) De eloping an a ea-wide sys em o coo dina ed
amp me e con ol., Technical Repo , T anspo a ion No hwes (O ganiza ion).
Yoshino, T., T. Sasaki and T. Hasegawa (1995) The a ic-con ol sys em on he hanshin
exp essway, In e aces,25 (1) 94–108.
Zhang, H. M. and S. G. Ri chie (1997) F eeway amp me e ing using a i icial neu al
ne wo ks, T anspo a ion Resea ch Pa C: Eme ging Technologies,5(5) 273–286.
Zhang, L. and D. Le inson (2005) Balancing e iciency and equi y o amp me e s, Jou nal
o T anspo a ion Enginee ing,131 (6) 477–481.
Zhang, M., T. Kim, X. Nie, W. Jin, L. Chu and W. Recke (2001) E alua ion o on- amp
con ol algo i hms, Cali o nia Pa ne s o Ad anced T anspo a ion Technology.
Zhao, D., X. Bai, F.-Y. Wang, J. Xu and W. Yu (2011) Dhp me hod o amp me e ing
o eeway a ic, IEEE T ansac ions on In elligen T anspo a ion Sys ems,12 (4)
990–999.
80
Fai ness in Ramp Me e ing Feb ua y 2025
A Appendix & Addi ional Da a
Table 12: P esence o Ramp Me e ing Wo ldwide
S a e Coun y Ci y Yea Ci a ion
No h Ame ica US Illinois 1963 Mizu a e al. (2014a)
Eu ope
UK Bi mingham 1986 Agency (2010)
Sweden S ockholm 2003 T a ik e ke (2019)
Ne he lands Ams e dam 1989 Taale e al. (1994)
Ge many Duisbu g, Rhine-Ruh egion 1996 Uni e si ä Duisbu g-Essen (n.d.)
F ance Pa is 1996 Haj-Salem e al. (2001)
Swi ze land Zu ich 2022 Swi ze land (2022)
Tu key Is anbul 2016 Municipali y (n.d.)
Oceania Aus alia Melbou ne 1971 Johnson and Bajeno (2018)
New Zealand Auckland 1983 T anspo (n.d.)
Asia
Is ael Tel A i Commission (n.d.)
China Suzhou 2022 ThePape (2022)
Taiwan 1982 Ins i u e o T anspo a ion, MOTC (2023)
A ica Sou h A ica Johannesbu g and Eku huleni 2006-2009 F eeway (2015)
Table 13: Exi In e al A e age Flow ( eh/h)
Con ol JE2 JE3 JE4 JE4b JE5
NoCon ol 1385 1106 1878 1411 1445
ALINEA 1181 1161 1559 1345 1422
UpS eamALINEA 1125 1167 1538 1292 1388
EqALINEA 1093 1125 1028 1058 1351
A e age 1096 1140 1501 1277 1401
81
Fai ness in Ramp Me e ing Feb ua y 2025
Table 14: Compa ison o Ramp Me e ing Algo i hms
Types o Algo i hms Pa ame e s Si e o De ec o S eng hs Limi a ion
Demand Capaci y D-C occupancy Ups eam
Downs eam
Measu ed om ups eam
low and downs eam occu-
pancy
The minimum me e ing a e is
used when downs eam occu-
pancy exceeds a c i ical alue
(Oc )
Demand Capaci y INRETS capaci y Downs eam
Mul i-de ec o s = 3
Measu ed om ups eam
low and downs eam occu-
pancy
Fo ee- lowing condi ions and
hea y conges ion
ALINEA algo i hm occupancy Downs eam
Measu ed om downs eam
occupancy
C i ical occupancy is mo e eli-
able han elying only on capac-
i y alue
RWS algo i hm occupancy Ups eam
Measu ed om ups eam
low
Simila o he D-C algo i hm
Pe cen occupancy algo i hm D-C capaci y Ups eam
Measu ed om ups eam
low and occupancy
Acqui ed om he demand ca-
paci y (D-C) algo i hm
MALINEA occupancy Downs eam
Measu ed om occupancy
ups eam and downs eam
In he ups eam me ge sec ion,
ALINEA was unable o educe
conges ion. Choosing he op-
imum posi ion o he down-
s eam de ec o s’ s a ion is di -
icul
FL-ALINEA occupancy Downs eam
Measu ed om downs eam
low and occupancy
Choosing he op imal posi ion
o he downs eam de ec o s’
s a ion is di icul
Speed-Occupancy algo i hm occupancy and speed Ups eam
Measu ed om speed and
occupancy o ups eam
T a ic condi ions could be e-
lec ed by bo h speed and occu-
pancy pa ame e s
ANCONA algo i hm speed Downs eam
Measu ed om speed o
ups eam
Unable o de ec conges ion
when loca ed downs eam om
an ac i e bo leneck
Local ALENA occupancy Downs eam = 2
Measu ed occupancy om
downs eam
Maximizes mainline h oughpu
by main aining desi ed occu-
pancy
He o ALENA occupancy Ups eam
Downs eam
Measu ed occupancy om
ups eam and downs eam
Di icul o es ima e down-
s eam de ec o low d op when
he mainline is close o capaci y
82

Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 19: Toy model Occupancy Va ia ion J9.
Sou ce: Own sou ce
83
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 20: Toy model Occupancy Va ia ion J10.
Sou ce: Own sou ce
84
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 21: Toy model Occupancy Va ia ion J11.
Sou ce: Own sou ce
85
Fai ness in Ramp Me e ing Feb ua y 2025
Figu e 22: Toy model A e age Speeds J9.
Sou ce: Own sou ce
Figu e 23: Toy model A e age Speeds J10.
Sou ce: Own sou ce
86