1
DERIVING TRAFFIC DATA FROM A CELLULAR NETWORK
Johan P. Widebe g*, Noelia Cace es, F ancisco G. Beni ez
T anspo a ion Enginee ing
ESI- Uni e si y o Se ille
Descub imien os s/n, 41092 Se ille
Spain
*[email p o ec ed]
ABSTRACT
Acqui ing high quali y o igin des ina ion in o ma ion o he ehicle a ic in a geog aphic
a ea is a edious and cos ly ask. T adi ional me hods a e expensi e, ime-consuming and
gene ally only p esen a snapsho o he a ic si ua ion a a ce ain ime. The echnique
de eloped in his pape exploi s he use o da a al eady a hand in a GSM ne wo k. Ins ead o
moni o ing ehicles lows, mobile phones lows a e measu ed and a e co ela ed o he a ic
low. This me hodology is based on he ac ha a GSM ne wo k always knows an es ima ed
posi ion o each e minal, e e ed o he loca ion a ea o he base s a ion ha p o ides
se ices o i . Fo a pilo s udy a GSM ne wo k simula o has been designed o gene a e a
syn he ic da abase wi h loca ion egis e s, which is hen p ocessed ma hema ically and
ans o med in o a ic da a. P ima y esul s show g ea po en ial o his me hod.
KEYWORDS
OD ma ix, a ic moni o ing, mobile phone acking, ne wo k simula ion, mobile phone
posi ioning.
INTRODUCTION
O igin-Des ina ion ma ices a e impo an o designing, analysing and enhancing oad
ne wo ks. These ma ices gi e in o ma ion abou ips o an a ea o in e es , p o iding an
es ima e o he numbe o ehicles a elling be ween poin s in his a ea o e a pe iod o ime.
An OD ma ix is di icul and o en cos ly o ob ain by adi ional me hods such as household
su eys o oadside moni o ing. In addi ion o he high cos s, hey a e sampling o da a is e y
ime-consuming and gene ally only p esen a snapsho o he a ic si ua ion a a ce ain ime.
In ecen yea s mobile phone se ices ha e become widesp ead all o e he wo ld. In o de o
p o ide se ices o mobile phone subsc ibe , mobile phone ne wo ks ha e an inhe en sys em
o iden i y he cu en loca ion o each mobile phone. Many applica ions will be a ailable in
he u u e exploi ing his loca ion in o ma ion. One o hese applica ions can be used in he
ield o su ace anspo a ion. The no el me hod p oposed in his pape uses in o ma ion
2
abou mobile phones loca ions o c ea e he a ic da a. I a mobile phone is p esen in a
ehicle and swi ched on, ha ehicle becomes a “p obe ehicle”. Then, mobili y in o ma ion
will be ob ained analyzing he loca ion o mobile phones on boa d o hem (p obe ehicles).
Figu e 1 - Vehicle wi h mobile phone
S a e o he A
A ai numbe o s udies ela ing o his ma e ha e been published in ecen yea s. Bolla [1]
p esen ed a model o es ima ing a ic by means o an algo i hm ha calcula es a ic
pa ame e s on he basis o elephone localiza ion da a. Lo ell ca ied ou a ious s udies [2]
[3] on he basis o anonymous da a on he posi ion o mobile phones wi h a iew o anspo
applica ions such as jou ney imes o speeds. We mus also e e o he in es iga ions by Sohn
[4] and Akin [5] on O-D ma ix calcula ions using simula ions o mobile phone da a, as well
as o he s udies ocussed on he e icacy o he echnique, such as ha by Cay o d [6], which
analyzed he main pa ame e s o be aken in o accoun , namely: p ecision, me e ing equency
and he numbe o localiza ions necessa y o achie e accu a e a ic desc ip ion.
Se e al pilo p ojec s ha e been unde aken o s udy his echnique, some o hem using he
ac ual localiza ion ca ied ou by GSM ope a o s when a mobile phone is simply swi ched [7]
on and o he s employing ce ain da a associa ed wi h speci ic si ua ions such as calls (da a
om phone bills) o hando e s [8] [9] [10]. The e ha e e en been p ojec s ha ha e
de eloped hei own so wa e o cap u e mobile phone signals [11] [12] [13]. All o hem
ha e ocussed on he use o localiza ion da a associa ed wi h mobile phones o ob ain
in o ma ion on ehicle mo emen s, while espec ing a he p i acy o he mobile phone use s.
MOBILITY MANAGEMENT IN CELLULAR NETWORKS
One o he main ea u es o mobile phone ne wo ks is he ac ha mobile s a ions ha e no
pe manen connec ion o he sys em. In addi ion, his sys em has o deal wi h ansi o y use s.
In o de o p o ide se ices o i s mobile subsc ibe s wi h sho delays and a low signalling
cos , he mobile phone ne wo k has o ack hem o be able o loca e he base s a ion (cell) in
which each phone is cu en ly loca ed. This is why loca ion a eas and cells a e in oduced. In
GSM e minology, he co e age a ea is di ided in o smalle a eas o hexagonal shape,
3
e e ed o as cells. A cell is de ined as he a ea in which a phone can communica e wi h a
ce ain base s a ion. A loca ion a ea (LA) is a geog aphic a ea co e ed by base s a ions
belonging o he same g oup. The iden i ie o LA associa ed o each g oup is called
“Loca ion A ea Iden i ie ” (LAI). O he iden i ie is he “Cell Iden i y” (CI) ha uniquely
iden i ies a cell in a loca ion a ea. A cell wi hin a GSM ne wo k is iden i ied by Cell Global
Iden i ie (CGI: LAI+CI).
Figu e 2 – Cells and loca ion a eas de ini ions
As he subsc ibe s a e ee o go e e ywhe e in he se ice a ea, i is ob ious ha hey will
en e and lea e cells. The e o e, loca ion in o ma ion mus be managed. The mobili y
managemen is he se o p ocesses used o allow he use s o be eached whe e e hey a e in
he ne wo k co e age a ea. One o he ope a ions in ol ed wi h mobili y managemen is he
loca ion upda ing. The loca ion upda ing e e s o p ocedu e whe eby he phone in o ms he
ne wo k abou i s cu en loca ion by means o a igge . The loca ion upda ing can be di ided
in wo s eps. The i s consis s o sending signalling messages om a mobile s a ion o upda e
i s loca ion. The second is he loca ion egis a ion, which consis s o a se o p ocesses which
upda es he ne wo k da abases, acco ding o loca ion messages which a e sen by he mobile
s a ions. This loca ion in o ma ion is upda ed and s o ed in he HLR and he VLRs. The
“loca ion upda ing” p ocedu e is ca ied ou when ce ain e en s a e p oduced, o ins ance a
phone mo es in o a cell which belongs o a new loca ion a ea o a ime associa ed wi h a
loca ion p ocess expi es. The loca ion in o ma ion, which is upda ed in his p ocedu e, is he
LAI and CI. These iden i ie s a e co esponded wi h he loca ion in o ma ion o he base
s a ion (cell) in which he phone has been egis e ed.
New al e na i e me hod
As ega ds he o e all opology o he ne wo k, i s co e age a ea will be ex ensi e in ci ies
and along he main elemen s o he anspo in as uc u e e.g. ailway lines, unk oads and
he like. In he case o a oad, he p esence o mobile phones in ehicles may be used o
es ima e eal ime a ic condi ions. As men ioned abo e, he GSM ne wo k p oduces an
es ima ion o a mobile phone posi ion when i is swi ched on wi hin he ne wo k’s co e age.
These es ima ed posi ions e e o he cell and loca ion a ea o he base s a ion in whe e each
mobile phone has been connec ed ( egis e ed) and hey a e upda ed and s o ed pe iodically in
he ne wo k da abases (HLR y VLR).
4
Figu e 3 - An example o ehicle ajec o y
I is possible o build ips by means o he ma ched phone posi ions combining he cells
whe e a mobile phone has been egis e ed and ca ying ou a map ma ching p ocess, om
which he o igin des ina ion ma ix can be in e ed. I is necessa y o bea in mind ha his
me hod is only s udy he ip i sel and no he a elle . To keep he pe sonal p i acy, he
mobile phone iden i y is enc yp ed using a unidi ec ional iden i ie hus, once enc yp ed,
canno be eco e ed.
PILOT STUDY
The aim o he pilo s udy is o show he easibili y o ob aining a ic coun s da a and o
upda e he OD ma ix om anonymous phones which a e on boa d he ehicles. Fo a i s
pilo s udy, a GSM ne wo k simula o has been designed and a syn he ic da abase was
gene a ed. The la e is hen p ocessed ma hema ically and u ned in o an o igin–des ina ion
ma ix and a ic coun s.
GSM Ne wo k Simula o
Cellula ne wo ks ha e a da abase ha eco ds he cu en loca ion o mobile phones o e
ime. The phone company has o know he loca ion o all mobile phones a any ime o
p o ide a good se ice o i s subsc ibe s. The simula o gene a es a syn he ic da abase wi h
he loca ion da a o mobile phones ha a e swi ched on du ing a day, emula ing a GSM
ne wo k da abase. The simula o emula es a co ido in an in e -u ban a ea co e ed by a
cellula ne wo k. This co ido has a se o cells and loca ion a eas which a e sp ead o e he
oad. Each cell co e s a oad link, along which he simula ed ehicle ypes a e a elling.
The simula o c ea es a se o ehicles wi h mobile phones on boa d acco ding o
ma hema ically and s a is ical c i e ia. The size o his se o ehicles is es ima ed using da a
om eal link coun s which a e si ua ed in he oad o in e es . The simula o manages he
mo emen s o his se o ehicles du ing a day and e en s which a e ela ed o he mobile
phones on boa d o he ehicles, o ins ance swi ching he phone on o o , making o
ecei ing calls, sms.... I also manages he “Loca ion Upda ing” p ocedu e when a egis e
e en is p oduced, e.g. a mobile s a ion mo es in o a cell which belongs o a new loca ion
a ea. Each ehicle has an assigned ip o ca y ou o e he co ido , an ins an o s a i s ip
and a numbe o on boa d mobile phones acco ding o i s occupan s. So ha , mobile phones
p esen in ehicles will be mo ed wi hin he cells ha p o ide he co e age o he co ido . I
i is p oduced some egis e e en while a phone is wi hin a cell, he mobile phone loca ion
will be eco ded in he syn he ic da abase o upda e i s loca ion, such as eal GSM “Loca ion
Upda ing” p ocedu e. Finally, he da abase will ha e s o ed he loca ion o all phones on
5
boa d o simula ed ehicles du ing a day. This loca ion da a is ela ed o he cell and loca ion
a ea in which a mobile phone has been egis e ed, like a eal GSM ne wo ks.
Table 1 – Example o “loca ion egis e ” gene a ed by he simula o
Time Enc yp ed ID phone LAI CI
2006-02-10 07:18:00 11303 2140155102 48751
2006-02-10 07:47:00 94499 2140156021 49833
2006-02-10 08:20:00 97404 2140155901 49301
I is necessa y o bea in mind ha he phone numbe o he subsc ibe name is no used o
ca y ou he analysis. Only i is necessa y an iden i ie o dis inguish be ween all mobile
phones. The echnical p o ec s he p i acy o mobile phone use s by enc yp ing phone
numbe s as soon as he acking da a is ecei ed. So ha , all phone numbe s in he da abase
a e enc yp ed using a unidi ec ional iden i ie o a oid any legal issues.
OD ma ix in e ence
To de elop his me hod, a syn he ic da abase wi h he loca ion eco ds du ing a day is used. In
his da abase, he eco ds a e al eady il e ed only conside s he mobile phones which a e
a elled h ough he a ea o in e es . Analysing he syn he ic da a gene a ed by he simula o ,
he mobile phones ha ha e been mo ed in o he cells ha co e he in e es ed oad is
eco e ed. Combining he cells whe e each mobile s a ion has been egis e ed, i is possible o
build a ip, om which he loca ion a ea o igin and des ina ion may be in e ed, which
s a ed a he momen when he i s egis e was eco ded and inished wi h he las egis e .
T a ic Coun s
Fo he pu poses o de e mining a ic low, he p ocess is based undamen ally on he ac
ha mobile e minals pe o m he whole “loca ion upda ing” p ocedu e ( egis a ion) when
hey en e a loca ion a ea wi h an iden i ie dis inc om he one hey had p e iously
eco ded. Fo his eason, i espec i e o whe he he e minal is eco ded in he ne wo k due
o any o he e en , he eco ds o a gi en phone will always be a ailable when i en e s a new
loca ion a ea, as long as i is swi ched on, whe he o no i is being used.
Figu e 4 - Example o Loca ion Upda ing on en e ing a new LA

6
Es ima ions o a ic olumes passing h ough hese bo de s a e ca ied ou by looking o he
momen s in which egis a ions ake place due o changes o LA, ha is o say momen s in
ime when he i s egis a ion occu s wi h a LAI di e en om ha p e iously eco ded. In
his way passage be ween a eas is moni o ed as i a ic coun s we e in ol ed.
Adjus men Fac o (Mobile phones pe Vehicle Equi alen s)
I is impo an o bea in mind ha his echnique gene a es da a associa ed wi h mobile
phones swi ched on o a pa icula ne wo k p o ide . Howe e , he main in e es is ge ing
ehicle a ic da a, no mobile phones da a. Consequen ly, i is used an adjus men ac o ha
ela es ehicles o mobile phones swi ched on. This ac o ( MPV) is used in he GSM da a
analysis o con e phones da a in o equi alen ehicles da a. I is de ined as equa ion (1):
)1(
1
·
ehicle
OpON
MPV
ehicle
OpONOpON CellPhones
VehiclesCellPhonesCellPhones =⇒=
The concep o es ima ing mobile phones pe ehicle equi alen (MPVE) is o calcula e
app oxima ely he numbe o mobile phones, which a e swi ched on and belong o a speci ic
ne wo k p o ide , displaced on boa d o each ehicle unde a speci ied a ea. This alue o
mobile phones pe ehicle is de ined using s a is ical pa ame e s o a mobile phone ma ke
om annual s udies published by he Telecommunica ions Ma ke Commission in Spain
(Mobile Phone Pene a ion, Ope a o Ma ke Sha e by Numbe o Subsc ibe s, ...), oge he
wi h o he a ic ehicula pa ame e s ( ehicle occupancy, p opo ion o ehicle ypes, e c).
RESULTS
The aim o his epo is o show he easibili y o using he da abase o mobile phones
loca ions o in e OD ma ix by means o phones. This da a could be ex ac ed di ec ly om a
mobile phone ne wo k bu o do ha coope a ion wi h a phone company would be necessa y.
Be o e en e ing any ag eemen s we ha e op ed o do a pilo s udy using a syn he ic da abase
gene a ed by a simula o .
The esul s ob ained om he analysis o his da abase p o ide an es ima e o he numbe o
phones a elling be ween poin s on a ne wo k o e a gi en pe iod o ime. The phones ips
a e hen u ned in o ehicles ips by an “adjus men ac o ”. These es ima ed ehicles ips
can be alida ed by means o he eal ones gene a ed by GSM Ne wo k simula o , since he
simula o also s o ed he eal ips in o de o make he compa ison wi h he es ima ed ones.
0
5
10
15
20
3:00 6:00 8:00 9:00 10:00 12:00 15:00 18:00 21:00 23:59
E o [%]
Bo de LA 1 - LA 2
Bo de LA 2 - LA 3
Bo de LA 3 - LA 4
E o a e age
Figu e 5 - T a ic Coun Es ima ion E o om simula ion o e each pe iod
7
0
5
10
15
20
25
00:00-6:00
00:00-9:00
00:00-12:00
00:00-15:00
00:00-18:00
00:00-23:59
(a)
E o [%]
0
1
2
3
4
5
E o [%]
LA 1
LA 2
LA 3
LA 4
LA 1LA 2LA 3LA 4
(b)
00:00 - 23:59
Figu e 6 - O-D Ma ix Es ima ion E o om simula ion o e di e en pe iods.
This e o is o igina ed p incipally by he “adjus men ac o ” used o ans o m phones ips
o ehicles ips, which is calcula ed using a e age s a is ical igu es e alua ed in one day. I
can be seen ha he co ec ion adjus men is almos pe ec , p oducing signi ican ly educed
le els o e o . Logically, his depends on he ac o employed, which is calcula ed using he
same a e age alues as hose used o de ine he s a is ical dis ibu ions modelling he
simula o ’s ehicles and mobile phones. Ob iously, calcula ion o he adjus men ac o will
no be so i ial when eal da a a e used, enabling mo e complex analysis o be unde aken.
O he ac o s also in luence in his es ima ion e o . One o hem is due o mobile phones ha
a e swi ched o du ing a whole. These phones p o ide e oneous in o ma ion because ha
ip is no de ec ed. Mo eo e , i a phone is swi ched on du ing a ip, he es ima ed ip is no
ei he he eal one because i is conside ed ha s a ed when he phone swi ched on bu no
when he ehicle began he mo emen .
Howe e , he educed e o le els ob ained in mul iple simula ions and analysis ha e
demons a ed ha i is possible o ob ain a ic da a om a cellula ne wo k, show he
po en ial o his echnique o collec a ic da a o ehicles using he loca ion da a o phones
on boa d o hem.
CONCLUSIONS
The use o mobile phones on boa d o ehicles such as measu ing ools (wi hou e ealing
con iden ial da a o a ec ing i s pe o mance) p esen s a ious ad an ages wi h ega d o
adi ional me hods. The main one lies in he possibili y o es ima ing a ic da a wi h low
deploymen cos , a oiding he need o ins alling addi ional de ices along he oad ne wo k o
he use o o he cos ly me hods such as a ic su eys. The ex ac ed da a come om he
same e minals ha mos people use nowadays, by he simple ac ha he phones a e
swi ched on in any zone wi h GSM co e age. Consequen ly, and bea ing in mind he ac ha
he p ocess is as and supposes o handle a widesp ead sample o people, his echnique based
on anonymous mobile phone da a ep esen s a majo e olu ion in a ic mobili y s udies.
8
REFERENCES
[1] Bolla R., Da oli F., Gio dano A. (2000). “Es ima ing oad a ic pa ame e s om
mobile communica ions”. In P oceedings 7 h Wo ld Cong ess on ITS, Tu in, I aly.
[2] Lo ell Da id J. (2001). “Accu acy o speed measu emen s om cellula phone ehicle
loca ion sys ems”. ITS Jou nal Vol. 6, No. 4, pp. 303-325.
[3] Smi h B ian L., Pack Michael L., Lo ell Da id J. and Se mons M. William (2001).
“T anspo a ion managemen applica ions o anonymous mobile call sampling”. In
P oceedings 11 h Annual Mee ing o ITS Ame ica, Miami, FL.
[4] Sohn K. (2004). Dynamic Es ima ion o O igin-Des ina ion lows Using Mobile phones
as P obes. SDI 2004-R-04. Depa men o U ban T anspo a ion. Seoul De elopmen
Ins i u e. Ko ea.
[5] Akin D. and Sisiopiku V.P. (2002). “Es ima ing O igin-Des ina ion Ma ices Using
Loca ion In o ma ion om Mobile phones”. P oceedings o 49 h Annual No h
Ame ican Mee ings o he Regional Science Associa ion In e na ional, Pue o Rico.
[6] Cay o d R. and Johnson T. (2003). “Ope a ional Pa ame e s A ec ing Use o
Anonymous Mobile phone T acking o Gene a ing T a ic In o ma ion”. 82 h TRB
Annual Mee ing.
[7] ITIS Holdings - P ess Releases. “Es imo ion Technology chosen o a ic collec ion in
An we p”. A ailable: h p://www.i isholdings.com/pd /i isp oximus.pd
[8] “Mobile phones gene a e a ic in o ma ion in B aban ”. Mobile T a ic Se ice
(MTS), de elopmen by LogicaCMG - h p://www.m s-li e.com
[9] Ra i C., Pulselli R. M., Williams S., F enchman D. (2005). “Mobile Landscapes: using
loca ion da a om mobile phones o u ban analysis”. En i onmen and Planning B –
Planning and Design.
[10] Thiessenhusen, K-U., Schä e R-P. and Lang T. (2003). “T a ic Da a om Mobile
phones: A Compa ison wi h Loops and P obe Vehicle Da a”. Ins i u e o T anspo
Resea ch, Ge man Ae ospace Cen e , Ge many.
[11] Ai sage - P ess Releases. “Using Cellula Technology o P o ide an In elligen T a ic
Roadmap”. A ailable: www.ai sage.com/asse s/Ai SageExecu i eSumma y.pd
[12] Ai sage - P ess Releases. “Mobile phones could keep you ou o a ic jams”.
A ailable: www.ai sage.com/asse s/Vi ginian_Pilo _4_27_05.pd
[13] Delcan Co po a ion – P ess eleases. “T a ele In o ma ion De i ed om Mobile
phones Begins o low om Inno a i e Public-P i a e Pa ne ship in Bal imo e”.
A ailable: h p://www.delcan.com/p od/index.php?id=295