You are on page 1of 27

Mobi l i t y Model i ng i n Thi r d Gener at i on Mobi l e

Tel ecommuni cat i on Syst ems


J. G. Mar k oul i dak i s, G. L. Ly ber opoul os, D. F. Tsi r k as, E. D. Sy k as
Nat ional Technical Universit y of At hens ( NTUA)
Depart ment of Elect rical and Comput er Eng.
9, Heroon Polyt echniou St r.
157 73, Zographou, At hens, GREECE
Tel. : + 30 1 772 2278, Fax: + 30 1 772 2530
e- mail: ymark@t elecom. nt ua. gr
Abst r act : I n mobile communicat ions, mobilit y modeling is involved in several aspect s relat ed t o
signaling and t raffic load analysis. I n t hird generat ion syst ems, t he influence of mobilit y on t he net work
performance ( e. g. , handover rat e) will be st rengt hened, mainly due t o t he huge number of mobile
users in conj unct ion wit h t he small cell size. I n part icular, t he accuracy of mobilit y models becomes
essent ial for t he evaluat ion of syst em design alt ernat ives and net work implement at ion cost issues. I n
t his paper, we propose t hree basic t ypes of mobilit y models, which are appropriat e for t he analysis of
t he full range of mobile communicat ions' design issues. The models provide different levels of det ail
regarding t he user mobilit y behavior. I n part icular: ( a) t he Cit y Area Model t races user mot ion at an
area zone level, ( b) t he Area Zone Model considers users moving on a st reet net work and ( c) t he St reet
Unit Model t racks user mot ion wit h an accuracy of a few met ers. The validit y of t he basic models for
mobile communicat ions design aspect s is highlight ed. Moreover, an "int egrat ed mobilit y modeling
t ool , which considers t he basic mobilit y models as component s is proposed, aiming at t he development
of a refined modeling approach. This is achieved by improving t he accuracy of t he input paramet ers of
each basic model, via t he exchange of some specific ( mobilit y relat ed) paramet ers among t he
component models. To j ust ify t he applicabilit y of t he proposed int egrat ed t ool for bot h t he analysis of
design aspect s and net work planning, indicat ive result s are present ed, derived from simulat ion- based
applicat ion examples of t he t hree basic mobilit y models.
1. Introduction
Third Generat ion Mobile Telecommunicat ion Syst ems ( TGMTS) , will be brought int o service t he early
years of t he next cent ury [ 1- 3] . Aiming at a mass market t elecommunicat ion syst em, t he TGMTS will
offer a plet hora of t elecommunicat ion services ( e. g. , voice, low and high bit rat e dat a, video) t o Mobile
Users ( MU) via a range of mobile t erminals
1
, operat ing in bot h public and privat e environment s ( office
areas, residences, t ransport at ion media, et c. ) . Due t o t he mass market nat ure, TGMTS should be
designed as a high capacit y syst em, able t o cope wit h t he envisaged overwhelming t raffic demands. To
achieve t his, a layered cell archit ect ure consist ing of macro- , micro- and pico- cells, has been adopt ed
[ 4] .
Compared t o second generat ion syst ems [ 5- 7] and apart from t he increased t raffic demands, t he
employment of locat ion management and handover procedures in a micro- cellular environment , in
conj unct ion wit h t he huge number of MUs
2
, will generat e a considerable mobilit y relat ed signaling load.
The increase of t he mobilit y relat ed signaling - apart from t he radio link- will have a maj or impact on
t he number of dat abase t ransact ions, const it ut ing t hus t he dat abase a possible bot t le- neck at t he fixed
net work side. Consequent ly, and given t he scarcit y of radio resources, met hods for signaling load
reduct ion are emerging for TGMTS [ 8- 21] . I t is obvious t hat opt imizat ion t echniques and efficient
net work planning algorit hms are crit ical issues, concerning t he overall TGMTS performance.
Mobilit y modeling is involved in t he analysis of: ( a) locat ion management relat ed aspect s ( locat ion area
planning [ 8, 12, 13, 18] , paging st rat egies [ 9, 11, 17] , et c. ) , ( b) radio resource management relat ed

1
The t erminals can range from basic low cost simple t ypes, t hrough hand port able devices and vehicle mount ed
speech devices up t o port able B- I SDN t erminals.
2
The expect ed user penet rat ion rat e ranges from 50% t o 70% of t he t ot al populat ion.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 2
aspect s ( mult iple access t echniques, channel allocat ion schemes, et c. ) [ 10, 14, 22] and ( c) propagat ion
relat ed aspect s ( fading, handover decisions, et c. ) [ 44, 45] . Evaluat ion st udies involve t he considerat ion
of user mobilit y behavior and t herefore, t he accuracy of t he result s ( and consequent ly t he conclusions)
heavily depends on t he assumed mobilit y models. Not e t hat t he accuracy of t he mobilit y models
involved in net work planning is highly desirable, since it may affect t he rat io of syst em capacit y vs.
net work implement at ion cost .
Several mobilit y modeling approaches ( simulat ion and analyt ical) can be found in t he lit erat ure.
Analyt ical models, based on simplifying assumpt ions, may provide useful conclusions regarding crit ical
net work dimensioning paramet ers [ 11, 17, 22- 30] . St udies on more realist ic analyt ical models, indicat e
t hat closed form solut ions can be derived for simple cases only ( e. g. , highways at free flow) [ 29, 30] . On
t he ot her hand, comput er simulat ion st udies [ 8- 10, 31, 32] consider more det ailed and realist ic mobilit y
models. Among t he disadvant ages of t he above models are: ( a) t he amount of t he required input
paramet ers, ( b) t he verificat ion of result s vs. real measurement s and ( c) t he required comput at ional
effort .
I n t his paper, t hree basic t ypes of mobilit y models which are appropriat e for t he full range of t he
TGMTS design issues ( e. g. , locat ion and paging area planning, handover st rat egies, channel
assignment schemes, et c. ) are int roduced. I n part icular:
The Ci t y Ar ea Model . I t consist s of a set of area zones connect ed via high capacit y rout es.
Candidat e out put paramet ers may include: t he user dist ribut ion per area zone vs. t ime, t he crossing
rat e per area zone, t he percent age of non- moving and moving users ( car passengers, pedest rians)
for each area zone vs. t ime, et c.
The Ar ea Zone Model . The model consist s of a st reet net work and a set of building blocks. I t may
be ut ilized for t he est imat ion of: t he pdf ( probabilit y dist ribut ion funct ion) of a user residence t ime in
an area zone, t he pdf of a user crossing t ime in an area zone, et c.
The St r eet Uni t Model . This model considers t hree st reet t ypes: ( a) highways, ( b) st reet s wit h
t raffic light ( s) cont rolled flow and ( c) high/ low priorit y st reet s. Candidat e out put paramet ers may
include: t he pdf of car densit y and car speed in a st reet segment , t he pdf of car residence t ime in a
st reet segment , et c.
Taking int o account t hat each model concent rat es on a specific range of design issues, we propose a
met hodological modeling approach, t he so- called I nt egrat ed Mobilit y Modeling Tool ( I MMT) , t hat
considers t he basic models as component s among which mobilit y- relat ed paramet ers can be
exchanged. The I MMT aims at :
t he improvement of t he accuracy of t he result s obt ained by each basic model,
t he validat ion of t he t heoret ical input assumpt ions and analyt ical models and
t he invest igat ion of t he effect of t he mobilit y model accuracy on design decisions.
The mat erial included in t his paper is organized as follows. I n sect ion 2, a general discussion on mobilit y
modeling is given, ident ifying candidat e input and out put paramet ers. Sect ion 3 st resses t he necessit y
for t he development of various mobilit y model t ypes for TGMTS. Sect ion 4 provides an overview of
t ransport at ion t heory approach in conj unct ion wit h t he t hree basic mobilit y models. I n sect ion 5, t he
I MMT approach is described. Sect ion 6 present s some indicat ive result s, which are derived from
simulat ion- based applicat ion examples of t he t hree basic mobilit y models ( Annexes A, B and C) . Annex
D provides a brief analysis on t he effect of mobilit y on t he user calling behavior. Finally, conclusions are
drawn in Sect ion 7.
2. Mobility Modeling in General
Mobilit y modeling at t empt s t o describe t he mobilit y behavior of an individual or a set of individuals.
Thus, a generic mobilit y model ( see Figure 1) can be described by defining:
The set of I nput Par amet er s Sin. This includes:
- A set of point s G, which represent s all possible locat ions.
- A set of individuals P, which const it ut es t he populat ion of t he model.
- A set of t ime inst ances T, i. e. , t he t ime period t he populat ions mobilit y is modelled.
The set of Out put Par amet er s Sout , consist ing of a set of funct ions. Each funct ion provides t he
locat ion of an individual p over t he set G for each t ime inst ance of t he set T:

{ }
S x x g p P g t G t T
out p p
: , , ( ) ,
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 3
where, gp( t ) is t he locat ion of t he individual p at t ime t over t he set of possible locat ions G.
Taking int o account t he nat ure of mobilit y, it is obvious t hat it is quit e difficult ( if not possible) t o eit her
measure or derive t he out put paramet ers via a set of funct ions/ met hods. This is mainly due t o t he huge
number of paramet ers t hat have t o be considered, regarding t he user mobilit y behavior, t he vehicular
t raffic condit ions, et c. To alleviat e t hese problems, t he simplified modeling approach described below
may be followed:
Popul at i on ( P ) : The populat ion is grouped int o classes e. g. , pedest rians, car passengers, working
people, st udent s, et c. , based on t he mobilit y behavior charact erist ics. Anot her possible simplificat ion
might be t he considerat ion of a represent at ive sample of t he t ot al populat ion.
Geogr aphi cal Ar ea ( G ) : The geographical area is organized int o regions wit h specific mobilit y
charact erist ics. I f so, t he locat ion of an individual is known wit h an accuracy of an area region, and
not wit h an accuracy of a specific point . Example cat egorizat ion of areas may be t he cit y cent er,
urban area, suburban area, rural area, et c.
Ti me Per i od ( T ) : The t ime period is divided int o t ime zones wit h specific mobilit y charact erist ics.
Example t ime zones might be t he rush hours, t he busy hour, t he night hours, t he weekends rush
hours, et c. Thus, depending on t he out put paramet ers, a specific t ime zone may be considered,
inst ead of long t ime periods ( e. g. , a week) .
Wit hin t his cont ext , t he set of out put paramet ers of t he simplified modeling approach is:

{ }
S x x g p P g t G t T
out p p
: , , ( ) ,
Not e t hat t he development of t he appropriat e mobilit y model should be based on a set of crit eria
regarding t he accuracy required in t erms of populat ion, space and t ime. Anot her crucial paramet er is
t he required effort . The effort may refer eit her t o t he collect ion of real measurement s or t o t he
demanded comput at ional cost .

Real Phenomena
Output Parameters
Measurements
Generic Mobility Model Generic Mobility Model
Functions and Methods
Input Parameters
Figure 1: Represent at ion of a Generic Mobilit y Model
3. The Necessity of Mobility Models in Mobile Communications
I n mobile t elecommunicat ions, service provision t o MUs is accomplished by t he employment of: ( a) t he
locat ion management procedures ( locat ion updat e, domain updat e, user regist rat ion, user locat ing,
et c. ) used t o keep t rack of t he user/ t erminal locat ion and ( b) t he handover procedure which allows for
t he cont inuit y of ongoing calls.
The performance of t he above procedures is influenced by t he user mobilit y behavior. Their applicat ion
direct ly affect s ( a) t he signaling load generat ed on bot h t he radio link and t he fixed net work ( e. g. ,
locat ion updat ing rat e, paging signaling load, et c. ) and ( b) t he dat abase queries load. Addit ionally, t he
handover procedure affect s t he offered t raffic volume per cell as well as t he Qualit y of Service ( QoS)
experienced by t he MUs ( e. g. , call dropping) . I n TGMTS, t he est imat ion of t he above paramet ers, which
are crit ical for net work planning and syst em design ( e. g. , locat ion and paging area planning, handover
st rat egies, channel assignment schemes, et c. ) urge for t he development of appropriat e mobilit y
models. Due t o diversificat ion of t he above issues, different mobilit y det ail levels are required. I n
part icular:
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 4
Locat ion Management Aspect s: Locat ion area planning, mult iple st ep paging st rat egies, dat a
locat ing st rat egies, dat abase query load, et c. [ 8, 9, 11- 13, 16- 21, 33- 35] . Locat ion management
relat ed issues require t he knowledge of t he user locat ion wit h an accuracy of a large scale area
( e. g. , locat ion or paging area) .
Radio Resource Management Aspect s: Cell layout , channel allocat ion schemes ( FCA, DCA) ,
mult iple access t echniques ( TDMA, CDMA) , syst em capacit y est imat ion, QoS relat ed aspect s,
signaling and t raffic load est imat ions, user calling pat t erns, et c. [ 10, 14, 22, 36- 43] . Radio resource
management relat ed aspect s require medium- scale area accuracy ( cell area) .
Radio Propagat ion Aspect s: Fading, signal st rengt h variat ion, handover decision algorit hms, et c.
[ 44, 45] . The analysis of radio propagat ion aspect s needs accuracy of a small- scale area
( comparable t o t he wavelengt h level) .
I n order t o t ackle wit h t he above ment ioned issues, t hree mobilit y model t ypes are proposed, namely,
t he cit y area, t he area zone and t he st reet unit model, which are described in det ail in t he following
sect ion.
4. Mobility Models for Mobile Communications
Wit hin t he cont ext of t he generic modeling approach present ed in Sect ion 2 and t aking int o account t he
requirement s relevant t o mobile communicat ions, t hree t ypes of mobilit y models are proposed. The
framework for t he development of t hose models is based on t ransport at ion t heory, an overview of
which is provided in t he following sect ion.
4.1. Overview of Transportation Theory Approach [46]
Transport at ion t heory aims at t he analysis and design of t ransport at ion syst ems ( e. g. , railways, st reet
net works, et c. ) . The basic issue t hat t he t ransport at ion t heory at t empt s t o resolve, is t he following:
Given a t ransport at ion syst em serving a cert ain geographical area, det ermine t he load t his syst em
should carry . The input framework which is ut ilized as a basis for t he development of t he relevant
models is described by t he following it ems:
Tr i ps: A t rip ( movement ) is charact erized by: ( a) t he purpose, ( b) t he end- point s ( originat ion-
dest inat ion) , ( c) t he t ransport at ion means ut ilized and ( d) t he rout e followed. Given a cert ain
geographical area, a t rip can be charact erized according t o it s end- point s locat ions as: ( a) int ernal
( bot h end- point s inside t he area) , ( b) out going ( originat ion inside t he area and dest inat ion out side) ,
( c) incoming ( t he opposit e of out going) and ( d) ext ernal ( bot h end- point s out side t he area) .
Ar ea Zones: Transport at ion t heory divides t he geographical area under st udy int o area zones. The
division is based on crit eria relat ed t o: ( a) t he populat ion densit y and ( b) nat ural limit s ( e. g. , rivers,
parks, highways, railway t racks, et c. ) . Not e t hat , t he t rip end- point s are considered wit h accuracy of
an area zone.
Popul at i on Gr oups: The populat ion of t he area under st udy is divided int o groups according t o
t heir mobilit y charact erist ics. Example groups are working people, resident ial users, st udent s, et c.
Movement At t r act i on Poi nt s ( MAP) : MAPs represent locat ions t hat at t ract t he populat ion
movement s and at which people spend considerable t ime periods. Examples are work places,
residences, t he shopping cent ers, et c. Each MAP charact erizes t he populat ion group t ype it at t ract s.
Ti me Zones: During a day t ime, it can be observed t hat t here are t ime periods during which cert ain
t ypes of movement s t ake place ( e. g. , movement s t owards work- places) and t ime periods where
cert ain populat ion groups reside t o cert ain MAPs ( e. g. , working hours, shopping hours, et c. ) . These
t ime periods are called t ime zones. Transport at ion t heory concent rat es on t he so- called rush hours,
where t he peak load occurs on t he t ransport at ion syst em under st udy.
Tr anspor t at i on Sy st ems Char act er i st i cs: A t ransport at ion syst em ( e. g. , a st reet net work, t he
urban buses net work, t he subway, et c. ) is charact erized by: ( a) it s capacit y, ( b) t he t rips it may
support and ( c) t he usage cost
3
.
The basic models used by t he t ransport at ion t heory are:
Trip Product ion and At t ract ion Models: The out put paramet ers of t hese models are t he number of
t rips produced and at t ract ed by each area zone. Example model is t he regression model [ 47] .

3
Cost is measured in t erms of t ime cost and money cost .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 5
Trip Dist ribut ion Models: The out put of t hese models is t he so- called originat ion- dest inat ion mat rix
OD( A
i
, A
j
) . Each element of t his mat rix equals t o t he number of t rips originat ed from area zone A
i
and dest ined t o area zone A
j
. Example model is t he gravit y model [ 48] .
Modal Split Models: The out put of t hese models is t he t ransport at ion mean an individual select s t o
perform a t rip wit h given end- point s. The maj or fact ors considered here are t he user annual income,
t he t ransport at ion mean usage cost [ 46] .
( Vehicular) Traffic Assignment Models: These models are used for t he est imat ion of t he probabilit y a
cert ain rout e is select ed, given t he t rip end- point s and t he st reet net work [ 49] . The crit eria ut ilized
here are t he rout e lengt h and t he usage cost .
4.2. City Area Model
A cit y area model describes t he user mobilit y and t raffic behavior wit hin a cit y area environment . The
need t o analyze t he user mobilit y behavior over large- scale geographical areas is raised by locat ion
management relat ed aspect s. Net work planning purposes, impose t he use of cit y area models
represent ing specific cit ies ( i. e. , based on geographical dat abases, demographic dat a and exist ing
t ransport at ion st udies) . On t he ot her hand, t ypical cit y area models are required for t he evaluat ion of
proposed syst em design alt ernat ives [ 8, 9] .
According t o t ransport at ion t heory, alt hough each individual cit y area exhibit s specific charact erist ics
( e. g. , populat ion dist ribut ion, dist ribut ion of MAPs, st reet net work, et c. ) , some generic charact erist ics
can be observed in most cont emporary cit ies. For example:
Cit ies are usually developed in such a way t hat densely populat ed areas ( urban areas) surround a
cit y cent er ( high densit y of work places and shopping cent ers) . While moving t owards t he cit y
edges t he populat ion densit y gradually decreases ( suburban and rural areas) .
The st reet net work support s t wo movement t ypes: ( a) radial ( i. e. , from t he cit y cent er t owards t he
edge of t he cit y and vice versa) and ( b) peripheral.
Applying t he met hodology described in sect ion 2, t he basic set of input and out put paramet ers of a
generic cit y area model is given below, while a t ypical applicat ion example model, adopt ed in various
st udies [ 8, 9] , is given in Annex A.

High Capacity Route
Figure 2: The Cit y Area Model
4.2.1. Input Parameters
Geogr aphi cal Ar ea: The geographical area covers t he whole cit y area, consist ing of a set of area
zones connect ed via high capacit y rout es ( see Figure 2) .
Area Zones: Combining t ransport at ion t heory aspect s and mobile t elecommunicat ion requirement s, it
seems reasonable t o assume t hat an area zone equals t o a net work area ( e. g. , macro- cell, local
exchange area, et c. ) .
High Capacit y Rout es: They represent t he most frequent ly select ed st reet s ( rout es) for t he support of
movement s bet ween different area zones.
Not e t hat areas out side t he cit y can also be modeled as area zones which at t ract s a relat ively low
number of t rips.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 6
Popul at i on: Mobile communicat ion syst ems focus merely on t he mobilit y behavior of MUs, and
t herefore, only t his relevant port ion of t he t ot al populat ion needs t o be considered. However, t aking
int o account t hat t hird generat ion syst ems are expect ed t o support high user penet rat ion rat es - similar
t o cont emporary fixed net works ( i. e. , in t he range of 70%) - it seems absolut ely inefficient t o simulat e
t he t ot al number of MUs
4
. I nst ead, a represent at ive sample of MUs seems sufficient .
The populat ion is divided int o MU groups, according t o t he mobilit y charact erist ics of t he individuals. For
each MU group, cert ain mobilit y and t raffic charact erist ics are assigned:
Mobilit y Behavior: The mobilit y behavior of an MU group is det ermined by t he t ype of movement s, t he
dist ribut ion of MAPs, t he movement init iat ion t ime, t he ut ilized t ransport at ion means, t he crit eria
according t o which a user select s t he rout e and t he t ime t he user spends at cert ain MAPs.
Traffic Behavior: Each MU group is assigned a set of paramet ers describing t he t raffic behavior of it s
members, i. e. , available services, call arrival rat es, call durat ion, et c. A det ailed analysis of t he MU
calling pat t erns is present ed in Annex D.
Ti me Per i od: As ment ioned above, t ransport at ion t heory concent rat es on t he so- called rush hours.
On t he cont rary, t elecommunicat ion syst ems design and dimensioning focus on t he well- known busy
hour. However, t here are problems, especially t he ones relat ed t o t he mobilit y relat ed signaling, which
require t he analysis of bot h t he high mobilit y hours ( rush hours) and t he busy hour.
4.2.2. Output Parameters
The out put paramet ers may include:
Amount of Locat i on Updat es for a given Locat ion Area ( LA) planning scheme. Various, eit her
st at ic or dynamic LA planning schemes have been proposed in t he lit erat ure [ 8, 12, 18] . The exist ing
analyt ical models [ 24- 27] , for t he est imat ion of an area ( LA in t his case) border- crossing rat e,
assume uniform user dist ribut ion, const ant average speed and fixed percent age of moving users. I n
short t erm, t heir accuracy can be considered as higher compared t o t he cit y area model ( where
only t he high capacit y rout es are considered) . However, t heir assumpt ions are not valid for long
t ime periods, since mobilit y condit ions in a cit y area are quit e dynamic. Moreover, in cert ain cases,
more complex mobilit y models are required e. g. , overlapping LAs [ 12] or LAs defined according t o
t he individual MU mobilit y behavior [ 8] .
Pagi ng Si gnal i ng Load in Mult iple St ep Paging St rat egies. Paging St rat egies which aim at t he
minimizat ion of t he paging signaling load have been st udied [ 9, 11, 17] . The crit ical paramet er here,
is t he probabilit y t hat a user is locat ed wit hin t he paging area ( a port ion of an LA) t hat t he paging
relat ed informat ion
5
indicat es [ 9] . Exist ing analyt ical models [ 11, 17] , based on simplifying
assumpt ions and det ailed simulat ion models [ 9] , provide useful result s for paging area
dimensioning.
Dat abase Quer y Load. Taking int o account t he solut ions proposed for t he ( dist ribut ed) dat abase
in TGMTS [ 19- 21] , it is necessary t o provide est imat ions regarding t he relat ive locat ion of t he
calling and called users, as well as t he relat ive locat ion of t he called user t o a cert ain dat abase node
( e. g. , Home Locat ion Regist er - HLR [ 5] ) . The cit y area model provides means t o est imat e t he
query load of a ( dist ribut ed) dat abase, which covers t he whole cit y area.
4.3. Area Zone Model
The evaluat ion of t he various radio resource management schemes requires t he knowledge of t he MU
locat ion wit h an accuracy of a micro- cell area. The model described in t his sect ion considers an area
zone ( see subsect ion 4. 2. 1) , consist ing of a set of building blocks and a st reet net work ( see Figure 3) ,
covered by several micro- cells. Similarly t o t he cit y area model, a specific area zone model can be
developed for net work planning purposes, while a t ypical area zone can be used for research reasons.
To derive a t ypical area zone model, regular- shaped building blocks and a regular st reet - net work
graph can be considered. The lat t er leads t o t he well- known Manhat t an grid, according t o which, an
area zone is represent ed by square shaped building blocks and ort hogonal grid st reet net work [ 10, 31] .

4
Assuming a large cit y of 10 millions inhabit ant s, t he number of mobile users will be in t he range of 7 millions ( 70%
penet rat ion) .
5
E. g. , t he paging area in which t he user was roaming during his most recent int eract ion wit h t he net work.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 7
Applying t he met hodology described in sect ion 2, t he basic set of input and out put paramet ers of a
generic area zone model is given below, while a t ypical applicat ion example model, adopt ed in [ 10] ,
is given in Annex B.
Figure 3: The Area Zone Model
4.3.1. Input Parameters
Geogr aphi cal Ar ea: The area consist s of a set of building blocks and t he corresponding st reet
net work.
Building Blocks: A building block is charact erized by: ( a) it s t ype ( resident ial, business, shopping
cent er, met ro st at ion, et c. ) and ( b) it s capacit y ( number of people) . The dist ribut ion of t he building
block t ypes wit hin a cert ain area zone depends mainly on t he area zone t ype ( cit y cent er, urban,
et c. ) .
St reet Net work: The st reet net work surrounds t he building blocks, forming in general a random graph.
The nodes of t he graph are t he crossroads. A st reet is charact erized by: ( a) it s size ( lengt h, number
of lanes) , ( b) t he average car speed, and ( c) it s orient at ion. Not e t hat pavement s are ( can be)
regarded as part of t he st reet net work.
Popul at i on: As j ust ified in subsect ion 4. 2. 1, only a sample of t he MUs populat ion needs t o be
modeled, wit hout affect ing t he accuracy level of t he out put paramet ers. For example, t he est imat ion of
t he offered t raffic load in radio resource management relat ed st udies, require t he considerat ion of ( at
least ) t he busy MUs.
Users may be cat egorized according t o t heir current mobilit y st at e as: not moving, pedest rians,
passengers, et c. For each MU group cert ain mobilit y and t raffic charact erist ics are assigned:
Mobilit y Behavior: The mobilit y behavior of an MU group is det ermined by t he mobilit y condit ions on
t he st reet net work. The basic random variables t hat describe t he mobilit y behavior of a moving MU
are:
t he t ime required t o cross t wo crossroads,
t he probabilit y of a user select ing a specific direct ion upon reaching a crossroad,
t he t ime an MU belongs t o a cert ain group ( not moving, pedest rian, car passenger, et c. ) ,
t he probabilit y t hat a user t ransit s bet ween different mobilit y st at es during a call [ 10, 31] and
t he MU t rip t ype ( e. g. , int ernal, ext ernal) .
I n general, an area zone is surrounded by ot her areas, t he incoming and out going MUs should also
be encount ered. To achieve a const ant average MU densit y wit hin an area zone ( an import ant
assumpt ion for t raffic analysis) , t he incoming and out going rat es should be equal ( see t hree- st ep
approach described in Annex B) .
Traffic Behavior: Each MU group is assigned a set of paramet ers describing t he t raffic behavior of it s
members ( i. e. , service profile, call arrival rat es, call durat ion, et c. ) . A det ailed analysis of t he MU
calling pat t erns is present ed in Annex D.
Ti me Per i od: Depending on t he purpose of t he st udy, eit her t he rush hours or t he busy hour should be
considered.
4.3.2. Output Parameters
An example set of out put paramet ers is present ed below.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 8
Si gnal i ng and Tr af f i c Rel at ed Par amet er s: offered t raffic load, resource ut ilizat ion fact or, handover
rat e, user residence t ime in a cell area, et c.
QoS Rel at ed Aspect s: call and handover blocking probabilit ies, call- dropping probabilit y, average
number of handovers per call ( and per user cat egory) , et c.
A set of analyt ical models has been proposed in t he lit erat ure for t he est imat ion of t he handover rat e,
as well as t he average number of handovers per call [ 22- 26, 30] . These models are based on
simplifying assumpt ions ( e. g. , exponent ial dist ribut ed user residence t ime in a cell) and may lead t o
generic conclusions ( e. g. , in [ 23] ) . I n layered cell archit ect ures, t he development of accurat e analyt ical
models is rat her complex, mainly due t o t he involvement of radio resource management aspect s ( see
st udies in [ 30] ) . Moreover, such models do not provide est imat ions for indoor- out door handover cases.
4.4. Street Unit Model
The st reet unit model describes t he mobilit y behavior of ( moving) MUs ( pedest rians, passengers, et c. )
wit h an accuracy of a few met ers. To develop such a model, a very det ailed analysis, regarding t he
car/ pedest rian mot ion and t he st reet t ype under any vehicular t raffic condit ions, is needed. The
analysis concerning t he car mot ion is present ed in Annex C, while t he basic set of input and out put
paramet ers of a generic st reet unit model follows.
4.4.1. Input Parameters
Geogr aphi cal Ar ea: The geographical area under considerat ion consist s of one or more st reet
segment s connect ed at crossroads ( see Figure 4) . A single segment ( st reet unit ) is charact erized by:
( a) it s lengt h, ( b) t he number of lanes and ( c) it s capacit y ( cars/ hr) . Concerning t he vehicular t raffic
flow cont rol, t he following st reet unit t ypes can be ident ified:
Highway: I t is charact erized by high average car speed, high capacit y, non- int errupt ed vehicular
t raffic flow and usually cont ains several lanes per direct ion. The lengt h of a highway may range from a
few up t o hundreds of kilomet ers.
Traffic Light Cont rolled Flow: Vehicular t raffic flow is cont rolled by t raffic light s. The t ime int erval t he
t raffic light remains in each st at e ( red, orange and green) is assumed const ant . The lengt h of such a
st reet unit may be in t he range of a building block side ( e. g. , 50- 200 m) .
Priorit ized Traffic Flow: I n t his st reet unit t ype, t he vehicular t raffic flow ( on crossroads) is cont rolled
by a set of driving rules, defined by t he use of specific signs ( e. g. , STOP, GIVE WAY, et c. ) The lengt h
of t his st reet t ype may be in t he range of a building block side ( e. g. , 50- 200 m) .
Popul at i on and Car Dr i v er Behav i or : The analysis of t he mobilit y paramet ers imposes t he
considerat ion of t he mobilit y behavior of any passenger/ pedest rian locat ed at t he st reet unit .
Pedest rians move at slow speeds ( 2- 5 km/ hr) , while t heir mot ion can be charact erized as cont inuous
( walking) or int errupt ed ( e. g. , shopping) . The behavior of a car driver can mainly be guided by t he
following rules:
- Minimizat ion of t he t ravelling t ime ( i. e. , based on rout e and speed opt imizat ion ) .
- Safe driving ( i. e. , t he car speed is limit ed by t he car densit y and t he st reet charact erist ics) .
Ti me Per i od: From t he descript ion of t he ot her mobilit y models, it is obvious t hat t he st reet unit model
should be analyzed under bot h rush hours and busy hour condit ions.
Car Speed v s. Car Densi t y : According t o t ransport at ion t heory, t here is st rong relat ion bet ween t he
car densit y and t he average car speed [ 46] . This is because safe driving requires a safet y dist ance
bet ween cars, which increases wit h t he car speed. I n general, t he average car speed on a st reet unit
may range from almost zero ( t raffic j am) up t o a maximum value, which is t he free flow speed
6
.
Not e t hat an import ant issue here, is t he assumed car arrival process. The applicat ion example of
Annex C follows a realist ic arrival process.
4.4.2. Output Parameters
Example out put paramet ers include t he pdf of t he following random variables:

6
The free flow speed corresponds t o very low car densit y condit ions, where t he car speed is not limit ed by t he safet y
dist ance but from t he st reet charact erist ics ( e. g. st reet widt h, driver visibilit y, et c. ) .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 9
The t ime an MU spends inside t he st reet segment ( residence t ime) .
The MU speed.
The number of MUs ( passengers, pedest rians) inside a cert ain st reet segment .
The st reet unit model can direct ly be used for propagat ion relat ed aspect s ( e. g. , slow fading analysis) .
Moreover, it is appropriat e for handover relat ed st udies [ 44, 45] , as well as for t he analysis of offered
t raffic load in micro- cells ( where t he number of users is quit e low and t he large number t heorems do
not provide realist ic est imat ions [ 30] ) .

Lane 1
Lane 2
Lane 3
V
m
a
x

=

v
3
Vmax = v2
Vmax = v4
Vmax = v5
Vmax = v1
Lane 4
Low Priority Street
Traffic Light Controlled Flow
Figure 4: The Generic St reet Unit Model
5. Mobility Models Integration
I n t his sect ion, we invest igat e t he int egrat ion of t he present ed basic mobilit y models, aiming at t he
derivat ion of a more realist ic modeling approach, which combines t he advant ages of t hese models.
Wit hin t his cont ext , t wo approaches can be envisaged:
Appr oach 1: The basic models are combined int o a single mobilit y model, having t he following
charact erist ics. The geographical area under st udy, covers a cit y area consist ing of a set of area zones
connect ed via high capacit y rout es. Each area zone is an area zone model, consist ing of a set of
building blocks and a st reet net work. Each st reet segment of t he area zone model is a st reet unit
model. I t is obvious t hat in order t o sufficient ly load t he st reet net work, t he whole ( cit y) populat ion
should be t aken int o account . Alt hough t his approach provides a high level of accuracy, t he
comput at ional effort ( simulat ion approach) and t he complexit y ( analyt ical approach) it int roduces is
probably overwhelming
7
.
Appr oach 2: The basic mobilit y models are regarded here as independent component s of an
I nt egrat ed Mobilit y Modeling Tool ( I MMT) , shown in Figure 5a. This approach succeeds in improving
t he input paramet ers accuracy of each basic model, by exploit ing mobilit y- relat ed out put paramet ers
derived by t he ot her models. This leads t o t he refinement of t he basic mobilit y models making t hus, t he
I MMT approach a powerful framework for t he analysis of mobile communicat ion syst ems relat ed

7
This was t he original reason for t he development of basic models, which alt hough individually aim at t he
invest igat ion of a limit ed set of design/ research aspect s, demand t olerable comput at ional effort .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 10
aspect s, at a reasonable comput at ional effort ( simulat ion approach) and t olerable complexit y ( analyt ical
approach) .
During t he refinement process of a basic mobilit y model ( see Figure 5b) , t hree met hods for t he
est imat ion of t he model out put paramet er( s) are evaluat ed: ( a) direct measurement s, ( b) simulat ion
models and ( c) analyt ical models. The aim of t he refinement process is t o increase t he accuracy of t he
est imat ed out put paramet er( s) at minimum cost . Cost in t his case may refer t o comput at ional cost
( e. g. , comput at ion t ime required by a simulat ion model) or t o real cost ( e. g. , corresponding t o t he cost
of measuring t he t arget paramet er( s) ) . I n t his cont ext , t he I MMT can be exploit ed for t he validat ion of
t heoret ical assumpt ions, t he evaluat ion of analyt ical models and t he invest igat ion of t he effect of t he
mobilit y modeling accuracy on syst em design decisions.

Real Phenomena
Output Parameters
Input Parameters
Street
Unit
Integrated Mobility Modelling Tool Integrated Mobility Modelling Tool
Area
Zone
City
Area
Mobility Related Parameters
Basic Mobility Model Refinement Process Basic Mobility Model Refinement Process
Input Parameters
Analytical
Models
Other Basic
Mobility Models
Simulation
Models
Validation of Results
Selection of Appropriate
Approach
Measurements
(a) (b)
Figure 5: ( a) The I nt egrat ed Mobilit y Modeling Tool ( I MMT) and ( b) t he Refinement Process of a Basic
Mobilit y Model Wit hin t he I MMT
An example list of paramet ers t hat could be exchanged among t he basic models is given below.
Ci t y Ar ea Model
The pdf of t he residence t ime wit hin an area zone ( measured by t he area zone model) .
The pdf of t he residence t ime in high capacit y rout es ( measured by t he st reet unit model) .
The durat ion of an int ernal t rip ( measured by t he area zone model) .
Ar ea Zone Model
The t rip t ypes, e. g. , int ernal, ext ernal, et c. ( measured by t he cit y area model) t hat can be ut ilized
so as t o enhance t he movement algorit hm of t he area zone model.
The MU moving st at es probabilit ies: ( a) not moving, ( b) pedest rians and ( c) car passengers
( measured by t he cit y area model) .
The pdf of: ( a) t he residence t ime in a st reet unit vs. st reet unit t ype, ( b) t he number of cars per
st reet unit t ype and ( c) t he car speed vs. st reet unit t ype ( measured by t he st reet unit model) .
St r eet Uni t Model
St at ist ics concerning more realist ic ( car) arrival rat es ( measured by t he cit y area and/ or t he area
zone model) .
St at ist ics regarding t he mot ion of pedest rians: cont inuous, int errupt ed ( measured by t he cit y area
model) .
6. Results
I n t his sect ion, indicat ive result s obt ained by t he applicat ion of t he example simulat ion models
present ed in Annexes A, B and C, are present ed. I n our st udy t he basic mobilit y models are det ailed
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 11
simulat ion ones, however, following t he I MMT approach one simulat ion model can be replaced by an
analyt ical as soon as t he analyt ical approach is proved t o provide result s of adequat e accuracy.
Ci t y Ar ea Model : Figure 6 present s t he percent age of MUs per area zone t ype ( cit y cent er, urban,
suburban and rural) vs. t ime. The result s indicat e t hat , apart from locat ion management st udies [ 8, 9] ,
t he cit y area model can provide a clear view of t he t raffic demand dist ribut ion over cert ain area zone
t ypes and it s ( t ime) variance. The lat t er result s can be ut ilized in DCA relat ed st udies [ 38- 40] .
Figure 7 depict s t he dist ribut ion of user movement s wit h respect t o an area zone of a cert ain t ype ( e. g. ,
cit y cent er) . Figure 8 illust rat es t he amount of area zone border crossings for out going users vs. t he
area zone for t he busy and t he rush hours. This t ype of result s, which are t he basis for LA planning
analysis [ 8] , can be compared t o t he crossing rat es provided by t he area zone model.
Figure 9 depict s t he percent ages of moving ( passengers, pedest rians) and not - moving users vs. t ime.
This figure also provides an example of how t he I MMT approach can be applied t o t he cit y area model.
I MMT columns refer t o result s obt ained by t he cit y area model, regarding cit y cent er and urban high
capacit y rout es as t raffic light cont rolled flow st reet unit s ( red/ green st at es durat ion equal t o 60/ 60)
while suburban and rural high capacit y rout es are regarded as highway st reet unit s. I n t his case t he
st reet unit model has been exploit ed so as t o measure t he pdf of t he t ime an MU spends in a high
capacit y rout e. The st reet unit model receives as an input t he st reet lengt h ( known from t he cit y area
geomet ry) as well as t he arrival rat e of cars as t his is est imat ed from t he init ial cit y area model
simulat ion ( t aking int o account t hat t he measurement s of t his model correspond t o a sample of t he
t ot al cit y populat ion) . As shown, t he I MMT approach slight ly alt ers t he result s. This is due t o t he fact
t hat t he cit y area model result s refer t o relat ively long t ime- periods and mainly depend on aspect s such
as geographical dist ribut ion of maps and number of performed t rips rat her t han t he specific pdf of t he
MU residence t ime in an area zone/ high capacit y rout e.

0
10
20
30
40
50
60
City Centre Urban Suburban Rural
Area Zone Type
P
e
r
c
e
n
t
a
g
e

o
f

P
o
p
u
l
a
t
i
o
n

p
e
r

A
r
e
a

Z
o
n
e

T
y
p
e
Initial
7:00-8:00
8:00-9:00
9:00-10:00
10:00-11:00
11:00-12:00
Figure 6: Percent age of Users per Area Zone Type vs. Day Time Period

0
10
20
30
40
50
60
City centre Urban Suburban Rural
Area type
P
e
r
c
e
n
t
a
g
e

o
f

m
o
v
e
m
e
n
t
s Internal
Outgoing
Incoming
External
Figure 7: Percent age of User Movement s vs. Area Type
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 12

1 3 5 7 9
1
1
1
3
1
5
1
7
1
9
2
1
2
3
2
5
2
7
2
9
3
1
0
50000
100000
150000
200000
250000
300000
350000
O
u
t
g
o
i
n
g

C
r
o
s
s
i
n
g

R
a
t
e

f
o
r

r
u
s
h

a
n
d

b
u
s
y

h
o
u
r
s

(
c
r
o
s
s
i
n
g
s
/
h
r
)
1 3 5 7 9
1
1
1
3
1
5
1
7
1
9
2
1
2
3
2
5
2
7
2
9
3
1
Area Zone
rush hour
busy hour
Figure 8: Out going Crossing Rat e vs. Area Zone for Rush Hours and Busy Hour

0
10
20
30
40
50
60
70
80
90
Initial 7:00-
8:00
8:00-
9:00
9:00-
10:00
10:00-
11:00
11:00-
12:00
Day Time Period
Not-Moving
Not-Moving (IMMT)
Pedestrians
Pedestrians (IMMT)
Passengers
Passengers (IMMT)
P
e
r
c
e
n
t
a
g
e

o
f

M
o
b
i
l
e

U
s
e
r
s
Figure 9: Percent ages of Moving ( passengers, pedest rians) and Not - Moving Users vs. Day Time Period
Ar ea Zone Model : Figure 10 depict s t he pdf of a busy MU cell residence t ime
8
, using: ( a) t he basic
area zone model ( Annex B) , ( b) t he model proposed by Hong and Rappaport [ 22] , ( c) t he I MMT
approach and ( d) t he exponent ial dist ribut ion
9
. To compare t he result ing pdfs, all cases refer t o a
common average MU cell residence t ime equal t o t he one of case ( a) . The assumed cell charact erist ics
appear in Figure B. 1 ( cross shaped, L= 185m) , while t he I MMT approach regards st reet s as t raffic light
cont rolled flow st reet unit s wit h red/ green st at es durat ion equal t o 60/ 60 sec. The st reet unit model in
t he I MMT approach, receives as an input t he st reet lengt h ( 185 m) while t he car arrival rat e is adj ust ed
so as t o achieve t he same average MU cell residence t ime as in case ( a) . As it can be observed, t he
exponent ial dist ribut ion is quit e good approximat ion for ( a) and ( b) cases. However, t he I MMT approach
indicat es t hat , in more realist ic models, t he exponent ial dist ribut ion approximat ion, alt hough applicable,
is not excellent .
Figure 11 depict s t he pdf of a busy MU cell crossing t ime
10
for t he above ment ioned cases ( a- d) . As it
can be observed, neit her t he Hong model nor t he exponent ial dist ribut ion provide a good approximat ion

8
I t represent s t he t ime period bet ween t he est ablishment of a call init iat ed inside a cell, and t he inst ance t he busy MU
crosses t he cell boundary.
9
A common assumpt ion in t he lit erat ure.
10
I t represent s t he t ime period bet ween t wo successive crossings of a cell area borders by a busy MU.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 13
for t he cases ( b) and ( c) . Case ( b) ( Annex B model) , which obviously approximat es a Gaussian
dist ribut ion, is closer t o t he realist ic approach of I MMT ( case ( c) ) .

0
0,02
0,04
0,06
0,08
0,1
0,12
2
,
5
1
2
,
5
2
2
,
5
3
2
,
5
4
2
,
5
5
2
,
5
6
2
,
5
7
2
,
5
8
2
,
5
9
2
,
5
1
0
2
,
5
1
1
2
,
5
1
2
2
,
5
1
3
2
,
5
1
4
2
,
5
Busy MU Residence Time (sec)
p
d
f
Hong
Basic Model
IMMT
Exp.Distr.
Figure 10: The pdf of t he Busy User Cell Residence Time for t he: ( a) Hong Model, ( b) Basic Area Zone
Model, ( c) I MMT Approach and ( d) Exponent ial Dist ribut ion

0
0,02
0,04
0,06
0,08
0,1
0,12
0,14
2
,
5
1
2
,
5
2
2
,
5
3
2
,
5
4
2
,
5
5
2
,
5
6
2
,
5
7
2
,
5
8
2
,
5
9
2
,
5
1
0
2
,
5
1
1
2
,
5
1
2
2
,
5
1
3
2
,
5
1
4
2
,
5
Busy MU Crossing Time (sec)
p
d
f
Hong
Basic Model
IMMT
Exp.Distr.
Figure 11: The pdf of t he Busy User Cell Crossing Time for t he: ( a) Hong Model, ( b) Basic Area Zone
Model, ( c) I MMT Approach and ( d) Exponent ial Dist ribut ion
Figure 12 depict s t he variat ion of t he call blocking and call dropping probabilit y
11
which derive from t he
simulat ion model of Annex B, for t he following cases:
i. Exponent ially dist ribut ed MU cell residence t ime where t he car densit y and average car speed are
t he ones assumed in Annex B.
ii. The I MMT approach, where t he st reet s of t he area zone model of Annex B represent t raffic light
cont rolled flow st reet unit s. The pdf of t he t ime a moving MU spends inside a st reet of t he area
zone model, is t he one measured by t he st reet unit model. The st reet unit model in t his case
receives as an input t he st reet lengt h ( 185 m) , t he red/ green st at es durat ion ( 60/ 60 sec) and t he
car densit y ( equal t o t he one assumed in Annex B as in case ( i) ) . Not e t hat in t his case a different
average car speed result ed ( equal t o 11. 035km/ hr) because Eq. B. 1 is not valid for t he t raffic light
cont rolled flow st reet unit .
iii. An alt ernat ive applicat ion of t he I MMT approach. The MU cell residence t ime in t he area zone model
of Annex B is assumed t o be exponent ially dist ribut ed ( as in case ( i) ) , however, since t he st reet s
represent t raffic light cont rolled flow st reet unit s, t he average car speed is t he one measured by
t he st reet unit model in case ( ii) . The car densit y on t he st reet s of t he area zone model is t he one
assumed in Annex B ( like in cases ( i) and ( ii) ) .

11
The call dropping probabilit y is t he probabilit y t hat a call is blocked eit her during call set up or during a handover
at t empt .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 14
As it can be observed, compared t o t he result s of cases ( ii) and ( iii) which are very close, case ( i)
provides an overest imat ion of call blocking and call dropping probabilit ies. The result s of case ( i) show
t hat t he speed- densit y relat ion affect s t he est imat ion of call blocking and call dropping probabilit ies. On
t he ot her hand, cases ( ii) and ( iii) indicat e t hat t he assumpt ion of exponent ially dist ribut ed MU cell
residence t ime is adequat e, provided t hat t he car densit y and average car speed have been accurat ely
est imat ed.
This st udy provides an example of how t he I MMT approach can be exploit ed so as t o verify t heoret ical
assumpt ions ( e. g. , validit y of t he exponent ially dist ribut ed MU cell residence t ime) and analyse t he
import ance of mobilit y relat ed input paramet ers ( e. g. , t he speed- densit y relat ion) for t he est imat ion of
cert ain QoS paramet ers t hat j udge t he syst em capacit y. Not e t hat , t he est imat ion of QoS paramet ers
can be significant ly affect ed in t he case t hat t he effect of mobilit y on t he user calling behavior is
encount ered ( see analysis in Annex D) .

0
0,005
0,01
0,015
0,02
0,025
0,03
0,035
0,04
0,045
0
,
4
5
3
0
,
4
5
7
0
,
4
6
0
,
4
6
4
0
,
4
6
7
0
,
4
7
1
0
,
4
7
4
0
,
4
7
8
0
,
4
8
1
0
,
4
8
5
0
,
4
8
8
0
,
4
9
2
0
,
4
9
5
0
,
4
9
9
0
,
5
0
2
0
,
5
0
6
0
,
5
0
9
0
,
5
1
3
0
,
5
1
6
0
,
5
2
User Penetration Rate
B
l
o
c
k
i
n
g

P
r
o
b
a
b
i
l
i
t
y
Call Blocking (i) Call Blocking (ii) Call Blocking (iii)
Call Dropping (i) Call Dropping (ii) Call Dropping (iii)
Figure 12: Call Blocking and Call Dropping Probabilit ies vs. User Penet rat ion Rat e for cases ( i) , ( ii) and
( iii) .
St r eet Uni t Model : The result s present ed in t his sect ion are based on t he model of Annex C regarding
a st reet unit model of lengt h L= 500m. The pdf of a car residence t ime
12
in a st reet unit is illust rat ed in
Figures 13 ( high and low priorit y st reet unit s) and 14 ( highway and t raffic light cont rolled flow st reet
unit s) . As it can be observed, t he pdf( s) are quit e different and do not seem t o fit t o some known pdf.
However, t he result s obt ained by t he st reet unit model provide means t o define approximat ing pdf( s) .
Figure 15 depict s t he pdf of t he number of cars in a t raffic light cont rolled flow st reet unit ( measured at
t he moment a new car ent ers t he st reet ) , while Figure 16 depict s t he pdf of t he car speed in t he same
st reet unit t ype. As it can be observed, t he pdf( s) approximat e t he Gaussian.
Figure 17 illust rat es t he average car speed in a t raffic light cont rolled flow st reet unit vs. t he car arrival
rat e and t he red/ green st at e durat ion. As shown, t he average speed decreases as t he proport ion of
red/ green st at es increases. This is because cars st op more frequent ly as t he durat ion of red st at e
increases ( against t he green st at e durat ion) and t herefore, t heir average speed is reduced.
Finally, Figure 18 depict s t he average car speed in high/ low priorit y st reet s ( for various t urning
probabilit ies ( TP)
13
) vs. t he car densit y in a high priorit y st reet . As it can be observed, t he average car
speed in low priorit y st reet s is rat her invariant , as t he car densit y in high priorit y st reet increases. This
is due t o t he fact t hat , as t he car densit y increases, t he dist ance bet ween cars decreases and t hus it is
more difficult for a car coming from a low priorit y st reet t o cross/ ent er t he high priorit y st reet . As a
consequence, t he average car speed decreases and t herefore, t he safet y dist ance bet ween a cross-
road and t he nearest car also decreases. On t he ot her hand, t he average speed is more st rongly
influenced by t he t urning probabilit y.

12
I t represent s t he t ime, during which a car moves along t his st reet unit t ype.
13
I t represent s t he probabilit y t hat a car ent ers a high priorit y st reet from a low priorit y one.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 15
Figures 17, 18 depict t he crit ical paramet ers t hat affect t he average car speed in t he cases t hat
vehicular t raffic flow is priorit ized or cont rolled by t raffic light s. Wit hin t his cont ext , t he result s
present ed in t hese figures provide means t o model t he relevant st reet unit t ypes in t he area zone
model.

5
5
.
5
6
1
.
5
6
7
.
5
7
3
.
5
7
9
.
5
8
5
.
5
9
1
.
5
9
7
.
5
1
0
3
.
5
1
0
9
.
5
1
1
5
.
5
1
2
1
.
5
1
2
7
.
5
1
3
3
.
5
0
0,05
0,1
0,15
0,2
0,25
p
d
f
5
5
.
5
6
1
.
5
6
7
.
5
7
3
.
5
7
9
.
5
8
5
.
5
9
1
.
5
9
7
.
5
1
0
3
.
5
1
0
9
.
5
1
1
5
.
5
1
2
1
.
5
1
2
7
.
5
1
3
3
.
5
Time (sec)
High Priority Str.
Low Priority Str.
Figure 13: The pdf of a Car Residence Time in: ( a) Low and ( b) High Priorit y St reet Unit

0
0,05
0,1
0,15
0,2
0,25
0,3
0,35
0,4
0,45
p
d
f
3 9
1
5
2
1
2
7
3
3
3
9
4
5
5
1
5
7
6
3
6
9
7
5
8
1
8
7
9
3
9
9
1
0
5
1
1
1
Time (sec)
Traffic Light Controlled Flow
Highway
Figure 14: The pdf of a Car Residence Time in: ( a) Highway and ( b) Traffic Light Cont rolled Flow St reet
Unit
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 16

0
0,01
0,02
0,03
0,04
0,05
0,06
1 4 7
1
0
1
3
1
6
1
9
2
2
2
5
2
8
3
1
3
4
3
7
4
0
Number of Cars
p
d
f
Figure 15: The pdf of t he Number of Cars in a Traffic Light Cont rolled Flow St reet Unit

3
,
6
7
,
2
1
0
,
8
1
4
,
4
1
8
2
1
,
6
2
5
,
2
2
8
,
8
3
2
,
4
3
6
3
9
,
6
4
3
,
2
0
0,05
0,1
0,15
0,2
0,25
p
d
f
3
,
6
7
,
2
1
0
,
8
1
4
,
4
1
8
2
1
,
6
2
5
,
2
2
8
,
8
3
2
,
4
3
6
3
9
,
6
4
3
,
2
Car Speed (Km/hr)
Figure 16: The pdf of t he Car Speed in a Traffic Light Cont rolled Flow St reet Unit

15
17
19
21
23
25
27
29
0,1 0,2 0,3 0,4
Car Arrival Rate (cars/sec)
A
v
e
r
a
g
e

C
a
r

S
p
e
e
d

(
k
m
/
h
r
)
30/60 60/60 60/30
Red/Green (sec)
Figure 17: Average Car Speed in a Traffic Light Cont rolled Flow St reet Unit vs. Car Arrival Rat e and t he
Red/ Green St at e Durat ion
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 17

0
5
10
15
20
25
30
35
6,49 12,61 21,97 26,57
Average Number of Cars in High Priority Street (cars)
A
v
e
r
a
g
e

C
a
r

S
p
e
e
d

(
K
m
/
h
)
High Prior. Str.
Low Prior. Str.(TP=0.2)
Low Prior. Str.(TP=0.3)
Low Prior. Str.(TP=0.4)
Figure 18: Average Car Speed in High/ Low Priorit y St reet vs. Car Densit y in t he High Priorit y St reet
( expressed as t he average number of cars on t he st reet )
7. Conclusions
I n t his paper, we have proposed a mobilit y modeling approach which cat ers for t he whole range of
design aspect s met in t hird generat ion mobile t elecommunicat ion syst ems ( e. g. , locat ion and paging
area planning, handover st rat egies, cell layout , channel assignment schemes, et c. ) . At a first st ep,
based on t he observat ion t hat t he analysis of various design aspect s requires different level of det ail
concerning t he user locat ion, t hree basic mobilit y models have been specified. I n part icular:
The ci t y ar ea model considers a cit y area as a set of area zones and high capacit y rout es. This
model is valid for t he analysis of locat ion management relat ed aspect s, where t he accuracy of t he
user locat ion is considered at t he level of a locat ion/ paging area.
The ar ea zone model considers an area zone as a set of st reet segment s and building blocks. The
model is valid for t he analysis of radio resource management relat ed aspect s, where t he user
locat ion is considered at t he level of a cell area.
The st r eet uni t model considers a set of st reet net work segment s. The model is valid for t he
analysis of radio propagat ion relat ed aspect s, where t he user locat ion is considered at t he accuracy
of a few wavelengt hs.
Taking int o account t hat each model focuses on a specific set of design issues, we have proposed a
met hodological modeling approach, t he so- called I nt egrat ed Mobilit y Modeling Tool ( I MMT) . The I MMT
approach considers t he basic models as independent component s among which mobilit y- relat ed
paramet ers can be exchanged. The result s obt ained, indicat e it s applicabilit y for t he validat ion of t he
t heoret ical input assumpt ions and t he result s of exist ing analyt ical models. I n t his cont ext , t he I MMT
approach cont ribut es in t he invest igat ion of t he effect of t he mobilit y model accuracy on design
decisions. Moreover, t he abilit y of t he I MMT t o represent any specific geographical area, const it ut es t he
t ool appropriat e for net work planning and t hus st resses t he relat ion bet ween mobilit y modeling and
net work implement at ion cost .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 18
ANNEX A
City Area Model Application Example
The cit y area under considerat ion ( see Figure A. 1) has a radius of 20 Km. Four area t ypes are
ident ified: cit y cent er, urban, suburban and rural area. The model consist s of 32 area zones ( 8 per cit y
area t ype) , 4 peripheral ( one per area t ype) and 4 radial high capacit y rout es. The populat ion is 6
millions inhabit ant s and t he MU penet rat ion rat e is 50% ( i. e. , t here are 3 millions of MUs roaming wit hin
t he cit y area) . I n our simulat ion t ool, a sample of 100000 MUs has been assumed.
Area Borders
Peripheral - Radial High Capacity Route
City Centre
Urban Area
Suburban Area
Rural Area
Figure A. 1: The Cit y Area Model Consist ing of Area Zones Connect ed via High Capacit y Rout es
Movement At t r act i on Poi nt s ( MAPs) : The following t ypes of MAPs are considered: ( a) residences,
( b) work places and ( c) ot her e. g. , shopping cent ers, parks, et c. Figure A. 2 present s t he assumed
dist ribut ion of MAPs over t he whole cit y area. Not e t hat wit hin a cert ain area t ype ( e. g. , urban,
suburban area) t he MAPs are uniformly dist ribut ed.
O
th
e
r
W
o
r
k
P
la
c
e
s
R
e
s
id
e
n
c
e
s
R
u
ra
l
S
u
b
u
rb
a
n
U
rb
a
n
C
e
n
te
r
0
10
20
30
40
50
%
Figure A. 2: The Dist ribut ion of MAPs over t he Cit y Area
MU Gr oupi ng based on t hei r Mobi l i t y Behav i or : As shown in Figure A. 3, MUs are grouped
according t o: ( a) t he mobilit y behavior t hey exhibit ( working people, resident ial users and high mobilit y
users) and t he ( b) t ransport at ion means t hey use ( privat e car, public t ransport at ion) .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 19
Working
People
Mobile Users
Residential High Mobility
Users
Private Car Taxi
Public
Transport.
Private Car
Taxi Private Car
Taxi
Public
Transport.
60%
25%
15%
35%
30% 70% 5%
60% 18%
16%
66%
Users
Figure A. 3: Cat egorizat ion of MUs According t o Their Mobilit y Behavior
Ti me Zones: The simulat ion t ime ranges from 7: 00 a. m. t o 12: 00 p. m. , including t hus bot h t he rush
hours ( 7: 00- 9: 00 a. m. ) and t he busy hour ( 11: 00 a. m. - 12: 00 p. m. ) .
I ni t i al MU Di st r i but i on over t he Ci t y Ar ea: MUs are init ially dist ribut ed over t he cit y area by
assuming t hat t he maj orit y ( 95%) is sit uat ed at t heir residences.
MU Movement s: The way MUs move depends on t he MU group, t he MAPs dist ribut ion, as well as t he
high capacit y rout es t opology. Upon ent ering t o a MAP, a residence t ime is assigned. When t his expires,
t he movement dest inat ion is select ed based on t he gravit y model [ 46, 48] , and a rout e is det ermined.
I n general, t he rout e from one area zone t o anot her, due t o t he high capacit y rout es t opology, is
discriminat ed int o t wo part s: ( a) radial and ( b) peripheral. I n our model, t he short est pat h approach is
applied: when an MU moves t owards an inner cit y area zone, first ly select s t he radial and t hen t he
peripheral direct ion. When an MU moves t owards an out er cit y ring, t he reverse order is followed i. e. ,
first peripheral and t hen radial.
Mov ement Al gor i t hm f or Wor k i ng Peopl e ( see Figure A. 4) : Working people init iat e a movement ,
at a moment in t ime uniformly dist ribut ed bet ween 7: 00 a. m. and 7: 30 a. m. Aft er a short period of
walking/ wait ing - depending on t he t ransport at ion mean t hat will be used- ( uniformly dist ribut ed
bet ween 5- 15 min. ) t he MU rides on a vehicle ( privat e car, t axi or a set public t ransport at ion means) .
Aft er anot her short period of walking, t he MU reaches his/ her work place. Some short dist ance
movement s around t heir work place may also be performed during t he working hours e. g. , lunch
t ime.
Mov ement Al gor i t hm f or Resi dent i al User s ( see Figure A. 4) : We assume t hat every resident ial
user init iat es a movement , at a moment in t ime uniformly dist ribut ed bet ween 7: 30 a. m. and 9: 30
a. m. We dist inguish t wo movement t ypes: ( a) t hose performed inside t he area zone where t he MU
residence locat es and ( b) t hose dest ined t o a different area zone. For movement s of t ype ( a) , MUs are
regarded as pedest rians for a t ime int erval uniformly dist ribut ed bet ween 0. 5 and 2 hours. For t ype
( b) movement s, MUs may use eit her t heir privat e car, a t axi or some public t ransport at ion mean. The
"residence t ime" wit hin t he dest inat ion area zone is uniformly dist ribut ed bet ween 0. 5 and 3 hours. I n
t his case, MUs are regarded as pedest rians visit ing banks, shops or ot her at t ract ion point s inside t his
area zone.
Mov ement Al gor i t hm f or Hi gh Mobi l i t y User s ( see Figure A. 5) : We assume t hat MUs belonging t o
t his cat egory are already in mot ion, having a dest inat ion when t he simulat ion st art s. Aft er having
complet ed t heir movement , t hey reside for a short period ( 0- 15 min. ) before t hey are assigned a new
dest inat ion. This procedure is repeat ed during t he whole simulat ion process.
Origin Pedestrian
Car
Taxi
Public Trans. Pedestrian Public Trans.
Pedestrian Destination
Figure A. 4. Movement Algorit hm for Working People and Resident ial users
Origination Destination
Car
Taxi
Figure A. 5. Movement Algorit hm for High Mobilit y Users
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 20
Mobi l i t y Condi t i ons: The average pedest rian speed is 5 km/ hr ( in [ 46] t he average pedest rian speed
is in t he range of 3km/ hr t o 5km/ hr) . The vehicle ( privat e car, t axi, bus) speed is based on Table A. 1.
Not e t hat , t he model could be furt her enhanced by t aking int o account t he relat ion bet ween car densit y
and speed on a cert ain st reet ( e. g. , based on t he model of Equat ion ( b. 1) ) .
Tr af f i c Condi t i ons
We consider t he arrival of bot h incoming and out going calls. The call arrival process is assumed t o be
Poisson. Table A. 2 present s t he call arrival rat e ( bot h incoming and out going) for each MU cat egory
during t hree day t ime periods: ( a) 7: 00a. m. - 9: 00a. m. , ( b) 9: 00a. m. - 11: 00a. m. and ( c) 11: 00a. m. -
12: 00p. m. I n addit ion, we present t he percent age of incoming and out going calls for each MU cat egory.
For high mobilit y users, t he rat e of incoming calls is assumed t o be higher t han t he corresponding
out going, because incoming calls are for bot h privat e and business communicat ions.
VEHICLE TYPE
AREA TYPE Public Transportation Media Car/TAXI
CENTRE 5 - 15 km/hr 10 - 20 km/hr
URBAN 10 - 30 km/hr 15 - 40 km/hr
SUBURBAN 30 - 70 km/hr 40 - 80 km/hr
RURAL 60 - 80 km/hr 60 - 100 km/hr
Table A. 1: The Range of Vehicle Velocit ies in Conj unct ion wit h t he Area Type
CALL ARRIVAL RATE (calls/MU/h)
MU Category 7:00-9:00 9:00-11:00 11:00-12:00 Incoming- Outgoing
Working People 0.5 2.0 3.5 50 % - 50 %
Residential Users 0.5 1.5 2.0 50 % - 50 %
High Mobility Users 1.0 2.0 3.5 70 % - 30 %
Table A. 2. Call Arrival Rat e ( calls/ MU/ h) for all MU Cat egories vs. Day Time
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 21
ANNEX B
Area Zone Model Application Example
Geogr aphi cal Ar ea: The geographical area is square shaped and represent s a t ypical cit y cent er area
( 4 km
2
) modeled as a Manhat t an grid i. e. , buildings are squares and st reet s form a regular grid. To
achieve a realist ic populat ion dist ribut ion, various t ypes of environment s have been defined, each
charact erized by: ( a) t he percent age of t he t ot al area it covers and ( b) t he corresponding populat ion
densit y ( see Table B. 1) .
Environment Coverage Percentage Population Density
(people/km
2
)
Busy Spots 30 % 150000
Business 41 % 75000
Domestic 10 % 9000
Streets 15 % (see Fig. B.1)
Other 4 % 6500
Table B. 1: Area Charact erist ics
The number of pedest rians and car passengers is calculat ed by t he following assumpt ions: ( a) 66. 67%
of t he st reet area is covered by cars, ( b) st reet s are bi- direct ional, ( c) t he average number of
passengers per car is 1. 5, ( d) t here are 2 buses per st reet segment , ( e) t he average number of
passengers per bus is 45 and ( f) pedest rians are uniformly dist ribut ed on t he pavement s ( see Figure
B. 1) . The result ing densit ies are 51000 passengers/ km
2
and 66000 pedest rians/ km
2
, where t he km
2
corresponds t o t he st reet area only and not t o t he whole geographical area.
Mobi l i t y Char act er i st i cs: Three MU groups are considered ( MUs are not allowed t o alt er cat egory
during a call) : ( a) users locat ed in buildings, ( b) pedest rians wit h a speed following t he Gaussian
dist ribut ion wit h mean value 5 km/ hr and variance 30%, and ( c) car passengers wit h a speed following
t he Gaussian dist ribut ion variance 30% and a mean value which is calculat ed by t he formula ( also
ut ilized in [ 29] ) :
v v
D
D
s f
jam

_
,

1 ( b. 1)
where,
vs: The average car speed on t he st reet ( km/ hr) .
vf: The free flow speed ( 60 km/ hr) .
D: The linear car densit y per lane ( cars/ km) .
Dj am: The linear car densit y per lane at t raffic j am ( cars/ km) .
I n our case it is obvious t hat D/ Dj am= 0. 667 ( since t he cars are assumed t o cover 66, 67% of t he st reet ) .
Thus, t he result ing average car speed is vs= 20 Km/ hr.
The MU movement direct ion may alt er only at crossroads. The same direct ion is kept wit h a probabilit y
of 0. 5, while t he user t urns left or right wit h a probabilit y of 0. 25 ( backward movement is not allowed) .
To model t he fact t hat MUs leave/ ent er t he area zone, we consider a Poisson process of incoming MUs
at a rat e equal t o t he rat e of out going which is est imat ed by a t hree- st ep approach:
1) Run t he simulat ion wit hout considering incoming users, and measure t he arrival rat e of out going
ones.
2) Run t he simulat ion wit h a generat or of incoming MUs at a rat e equal t o t he one measured in st ep
( 1) . I n t his case, a port ion of t he incoming MUs crosses t he area zone and as a result t he out going
rat e is st ill higher t han t he incoming.
3) Measure t he percent age of incoming MUs t hat cross t he area zone, and reduce t he rat e of incoming
MUs accordingly.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 22
L = 185 m
L (m)
Pedestrian
Pavement Building Block Street segment
Pavement
Lstr = 15 m
9 m
9 m
4 m
3 m
Figure B. 1: The Populat ion Dist ribut ion on t he St reet s of a Micro- Cell
Tr af f i c Char act er i st i cs: Traffic relat ed paramet ers are shown in Tables B. 2, B. 3 [ 10] . We assume a
Poisson call arrival process, and an exponent ially dist ribut ed call durat ion ( see Table B. 3) . The call
arrival rat e refers t o t he t ot al number of incoming and out going calls during busy hour condit ions.
Furt her improvement s of t he MU calling pat t erns can be based on t he analysis of Annex D ( e. g. , t he call
arrival rat e could differ bet ween business and resident ial users) . We assume t hat 65% of t he MUs
locat ed inside buildings are served by privat e wireless net works.
Regarding t he mult iple access prot ocol, t he one present ed in [ 50] has been adopt ed. The prot ocol is
capable of support ing t hree service classes, namely, circuit - mode voice, burst - mode voice and dat a, by
performing st at ist ical mult iplexing of connect ions of t he t hree classes at t wo different levels: ( a) t he
call- level ( for circuit - mode voice) and ( b) t he t alkspurt / message- level ( for burst - mode voice and dat a) .
I n our simulat ion, we assume t hat 70% of t he voice calls are of t he packet - mode t ype.
Regarding t he radio resource allocat ion, a Fixed Channel Allocat ion ( FCA) scheme is considered, while in
every micro- cell, 2 carriers wit h 96 full- duplex channels ( 96 slot s/ frame, FDD) are available [ 50] . A
micro- cell area equals t o t he area depict ed in Figure B. 1.
Service Type Call Arrival Rate
(Calls/User/Hr)
Mean Call
Duration (sec)
Voice 3,0 90
Data 2,78 50
Table B. 2: Call Arrival and Call Durat ion per Service Type
Ser vi ce Penet r at i on Rat e
Envi r onment Voi ce Dat a
Busy Spots 95 % 5 %
Business 99 % 70 %
Domestic 95 % 10 %
Street 98 % 2 %
Other 95 % 5 %
Table B. 3: Penet rat ion Rat e per Service and Environment
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 23
ANNEX C
Street Unit Model Application Example
I n t his annex, we describe a simulat ion t ool, which models all t ypes of st reet unit s i. e. , highways, t raffic
light cont rolled t raffic flow and priorit ized flow cont rol st reet s. For t he sake of simplicit y, our simulat ion
t ool considers a single lane st reet unit ( see Figure C. 1) . As far as t he behavior of a car driver is
concerned, t his is mainly guided by t he following rules:
- Minimizat ion of t he t raveling t ime ( i. e. , based on rout e and speed opt imizat ion) .
- Safe driving ( i. e. , t he car speed is limit ed by t he car densit y and t he st reet charact erist ics) .
L
(c) High-Low Priority Street
L
STOP STOP
High Priority Street
Low Priority Street
L
(b) Traffic Light Controlled Flow
S
L
(a) Highway
Figure C. 1: St reet Unit Models: ( a) Traffic Light Cont rolled Flow and ( b) High- Low Priorit y St reet
Based on t hose t wo rules, we built up an empirical law regarding t he way t he driver cont rols t he car
speed. This empirical law correlat es: ( a) t he car accelerat ion wit h t he dist ance bet ween a car and t he
foregoing car, ( b) t he maximum car speed on t he st reet unit and ( c) t he car accelerat ion- decelerat ion
charact erist ics. The empirical law is expressed by t he following formula:

v t
t
g k
v t c
v
c c
m
( )
exp
( )

+

_
,

1
]
1

1
]
1
1
1 1 ( c. 1)
where,
v( t ) : The car speed at t ime t ( km/ h)
gc: A const ant expressing t he car accelerat ing capabilit ies ( 20 m/ sec
2
) .
kc: A const ant expressing t he car decelerat ing capabilit ies ( assumed value 0. 6) .
c: An arbit rary const ant ( assumed value 0. 5 km/ hr) .
vm: The maximum safet y speed based on t he dist ance from and t he speed of t he preceding car
( km/ h) .
The vm is given by t he following empirical formula:
( )
v v
d t
d v t
v t c
m f
safe pre
p
pre

_
,

'

min ,
( )
( ( ))
( ) ( c. 2)
where,
d( t ) : The current dist ance bet ween t his car and t he preceding one ( km) .
vpr e( t ) : The current speed of t he preceding car ( km/ h) .
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 24
dsafe( v) : The safet y dist ance bet ween t wo cars t raveling at speed v ( km) .
p: An empirical const ant expressing t he sensit ivit y of t he driver t o t he speed changes of
t he preceding car ( assumed value 2) .
Finally, t he safe dist ance bet ween t wo cars t raveling at speed v, is given by t he following formula:
d v a v d
safe
( )
min
+ ( c. 3)
where,
: a const ant expressing t he t ime it t akes a driver t o st op t he car by t he moment t he preceding car
st art s decelerat ing ( t he assumed value is 0. 7 sec) .
dmin: t he minimum dist ance bet ween not - moving cars ( 0. 5 m) . Not e t hat t he average car lengt h is
assumed equal t o 5m.
The above ment ioned empirical law describes t he behavior of an individual driver in a st reet unit and
can be ut ilized for simulat ing t he car mot ion. To adopt realist ic values for t he paramet ers appearing in
t he above formulas, we have performed a series of experiment s. I n t hese experiment s, t he speed of a
single car j ust ent ering t he st reet unit wit h an init ial speed set t o zero, is analyzed for t he following
cases:
No ot her car on t he st reet ( free flow) .
One car at a dist ance of 30m t raveling wit h 10, 40 or 60 km/ hr.
One not moving car at a dist ance of 100m.
The experiment result s for t his set of represent at ive cases are present ed in Figure C. 2.
0
10
20
30
40
50
60
70
80
0,08 2,08 4,08 6,08 8,08 10,08 12,08 14,08 16,08 18,08
Time (sec)
C
a
r

V
e
l
o
c
i
t
y

(
k
m
/
h
r
)
Stop at 100m
Free Flow
Prec.Car 60km/hr
Prec.Car 40km/hr
Prec.Car 10km/hr
Figure C. 2: The Behavior of t he Empirical Model of t he St reet Unit Model. I t depict s t he car speed vs.
t ime for t he following cases: ( a) free flow ( no preceding car exist s) , ( b) preceding car at 30m wit h
speed 60km/ hr, ( c) preceding car at 30m wit h speed 40km/ hr, ( d) preceding car at 30m wit h speed
10km/ hr, ( e) preceding car st opped at 100m.
The st reet unit model has been ut ilized so as t o analyze all t ypes of st reet t ypes: ( a) highway, ( b)
t raffic light and ( c) high/ low priorit y st reet s. To achieve a more realist ic car arrival process, our model
considers t hree st reet unit s is series, where t he arrival process of cars in t he first st reet is Poisson
( default rat e: 0. 2 cars/ sec) while measurement s are performed in t he t hird st reet unit .
I n t he case of t raffic light , t he greed/ orange/ red st at es durat ion is 60/ 3/ 60sec, respect ively. I n t he case
of t he high/ low priorit y st reet s, t he probabilit y of a car in a low priorit y st reet ent ering a high priorit y
st reet is 0. 2. Cars in low priorit y st reet s, decide t o ent er/ cross t he high priorit y st reet when t he
dist ance bet ween t he cross- road and t he nearest arriving car on t he high priorit y st reet is safe. The
minimum safe dist ances for ent ering/ crossing t he st reet are assumed t o be a funct ion of t he speed Vcar
of a car moving in high priorit y st reet : cross: 1. 2* Vcar, ent er: 2* Vcar.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 25
ANNEX D
Mobile Users Calling Behavior
An import ant issue influenced by user mobilit y concerns t he MU calling behavior expressed by t he t he
incoming/ out going call arrival rat e and t he average call durat ion. From fixed net works it is well- known
t hat different calling behavior charact erizes business and resident ial users. I n mobile communicat ion
syst ems, different calling pat t erns can be ident ified for moving and not moving users. For example,
short er call durat ion is expect ed for car drivers compared t o not moving users.
The est imat ion of t he t raffic- relat ed paramet ers is subj ect t o t he following assumpt ions:
Not Mov i ng User s: Est imat ions can be based on relevant est imat ions from fixed net works.
Mov i ng User s: Est imat ions can be based on t he following assumpt ions:
The rat e of out going calls of a specific service t ype depends on t he user mobilit y class ( e. g.
pedest rian, car passenger, et c. ) . This is due t o t he fact t hat t he MU class affect s t he convenience of
a user t o init iat e calls. E. g. , compared t o a user sit uat ed in his office a pedest rian will normally
init iat e a lower number of voice calls and an even lower number of ( if not any) fax calls.
The rat e of incoming calls does not depend on t he user mobilit y class, since t he calling MU in general
ignores t he current moving st at e of t he called MU.
The call durat ion is st rongly affect ed by t he user mobilit y class. This is due t o t he fact t hat t he
mobilit y class det ermines t he user convenience for making longer calls ( e. g. , short er call durat ion
are expect ed for met ro passenger compared t o a privat e car passenger) .
The convenience t o communicat e ( by means of call init iat ion and call durat ion) is assumed t o be
affect ed by t he user mobilit y class, according t o t he hierarchy list ed in t able D. 1. Not e t hat t he higher
t he posit ion in t he hierarchy t he higher t he corresponding value is expect ed.
Cal l I ni t i at i on Cal l Dur at i on
Not moving business Not moving residential
Not moving residential Not moving business
Car passenger Car passenger
Public transportation passenger Pedestrians
Pedestrians Public transportation passenger
Table D. 1: The Hierarchy of User Mobilit y Classes
REFERENCES
[ 1] S. T. S. Chia, "The Universal Mobile Telecommunicat ion Syst em", I EEE Communicat ions Magazine, Vol. 30, No.
12, Dec. 1992, pp. 54- 62.
[ 2] J. Rapeli, UMTS: Target s, Syst em Concept , and St andardisat ion in a Global Framework , I EEE Personal
Communicat ions, Feb. 1995, pp. 20- 28.
[ 3] E. Buit enwerf, G. Colombo, H. Mit t s and P. Right , UMTS: Fixed Net work I ssues and Design Opt ions , I EEE
Personal Communicat ions, Feb. 95, pp. 30- 37.
[ 4] J. Sarnecki, C. Vinodrai, A. Javed, P. O' Kelly and K. Dick, "Microcell Design Principles", I EEE Commun. Mag. , Vol.
34, No. 4, April 1993.
[ 5] M. Mouly and M- B. Paut et , The GSM Syst em for Mobile Communicat ions , published by t he aut hors, Palaiseau,
France, 1992.
[ 6] M. Rahnema, " Overview of t he GSM Syst em and Prot ocol Archit ect ure" , I EEE Commun. Mag. , Vol. 31, No. 4,
April 1993, pp. 92- 100.
[ 7] ETSI / RES- 03( 91) 61 "DECT Syst em Descript ion Document ", Version 3. 4, June 1992.
[ 8] J. G. Markoulidakis, G. L. Lyberopoulos, D. F. Tsirkas and E. D. Sykas, Evaluat ion of Locat ion Area Planning
Scenarios in Fut ure Mobile Telecommunicat ion Syst ems , Wireless Net works Journal, Vol. 1, No. 1, Jan. 1995,
pp. 17- 29.
[ 9] G. L. Lyberopoulos, J. G. Markoulidakis, D. V. Polymeros, D. F. Tsirkas and E. D. Sykas, I nt elligent Paging St rat egies
for Third Generat ion Mobile Telecommunicat ion Syst ems , I EEE Trans. on Vehic. Tech. , Vol. 44, No. 3, Aug.
1995, pp. 543- 554.
[ 10] G. L. Lyberopoulos, J. G. Markoulidakis and M. E. Anagnost ou, "The impact of evolut ionary cell archit ect ures on
handover in fut ure mobile t elecommunicat ion syst ems", Proc. I EEE Vehic. Tech. Conf. ' 94, St ockholm, June
1994, pp. 120- 124.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 26
[ 11] J. G. Markoulidakis and E. D. Sykas, " Performance Bounds of a Mult iple St ep Paging St rat egy" , I nt ernat . Journal
of Wireless I nformat ion Net works. , Vol. . 2, No. 3, 1995, pp. 133- 147.
[ 12] S. Okasaka, et . al. , " A new Locat ion Updat ing Met hod for Digit al Cellular Syst ems" , 41st VTS Conference, St .
Louis, Missouri, 19t h- 22t h May 1991, pp. 345- 350.
[ 13] Behrokh Samadi and Wing S. Wong, " Opt imisat ion Techniques for Locat ion Area part it ioning", 8t h I TC specialist
Seminar on Universal Personal Telecommunicat ion, Geneva, Oct . 1992, pp. 141- 147.
[ 14] R. Beck and H. Panzer, " St rat egies for Handover and Dynamic Channel Allocat ion in Microcellular Mobile Radio",
I EEE VT Conf. Proc. , paper 5. 4, San Francisco, CA May 1989.
[ 15] S. T. S. Chia, "A Handover Prot ocol for a Mixed Cell Syst em" , 6t h I EE I nt er. Conf. on Mobile Radio Personal
Commun. , Warwick, RI , Dec. 1991, pp. 225- 232.
[ 16] S. T. S. Chia, " Locat ion Regist rat ion and Paging in a Third Generat ion Mobile Syst em" , BT Technology Journal,
Vol. 9, No 4, Oct . 1991, pp. 61- 68.
[ 17] S. Madhavapeddy, Opt imal Paging in Cellular Mobile Telephone Syst ems , Proc. I TC 14, 1994, pp. 493- 502.
[ 18] H. Xie, S. Tabbane, D. Goodman, Dynamic Locat ion Area Management and Performance Analysis , Proc. I EEE
Vehic. Tech. Conf. , Secaucus, NJ, 1993, pp. 536- 539.
[ 19] S. Mohan, R. Jain, Two User Locat ion St rat egies for Personal Communicat ions Services , I EEE Personal
Communicat ions, 1994, pp. 42- 50.
[ 20] S. Tabbane, Evaluat ion of an Alt ernat ive Locat ion St rat egy for Fut ure High Densit y Wireless Communicat ions
Syst ems , WI NLAB Tech. Report , July 1992.
[ 21] J. Wang, A Fully Dist ribut ed Locat ion Regist rat ion St rat egy for Universal Personal Communicat ion Syst ems ,
I EEE Journ. of Select . Areas in Comms. , Vol. 11, No. 6, Aug. 1993, pp. 850- 860.
[ 22] D. Hong and S. S. Rappaport , " Traffic Model and Performance Analysis for Cellular Mobile Radio Telephone
Syst ems wit h Priorit ized and Nonpriorit ized Handoff Procedures" , I EEE Transact ions on Vehicular Technology,
Vol. VT- 35, No. 3, Aug. 1986 pp. 77- 92.
[ 23] S. Nanda, " Telet raffic Models for Urban and Suburban Microcells: Cell Sizes and Handoff Rat es", I EEE Trans. on
Vehic. Tech. , Vol. 42, No. 4, Nov. 1993, pp. 673- 682.
[ 24] J. G. Markoulidakis, E. D. Sykas, " Model for locat ion updat ing and handover rat e est imat ion in mobile
t elecommunicat ions", I EE Elec. Let t ers, Vol. 29, No. 17, Aug. 1993, pp. 1574- 1575.
[ 25] G. Morales- Andres and M. Villen- Alt amirano, "An approach t o modeling subscribers mobilit y in cellular radio
net works", Forum Telecom 87 Geneva, 1987.
[ 26] W. Mende, " On t he Hand- Over rat e in Fut ure Cellular Syst ems" , I EEE Vehic. Tech. Conf. , 1988, pp. 358- 361.
[ 27] E. Alonso, K. S. Meier- Hellst ern and G. P. Pollini, "I nfluence of Cell Geomet ry on Handover and Regist rat ion Rat es
in Cellular and Universal Personal Telecommunicat ions Net works", 8t h I TC specialist Seminar on Universal
Personal Telecomm. , Geneva, Oct . 1992, pp. 261- 270.
[ 28] G. Mont enegro, et . al. , " Time Dependent Analysis of Mobile Communicat ion Traffic in a Ring- Shaped Service
Area wit h Non- uniform Vehicle Dist ribut ion", I EEE Trans. on Vehic. Tech. , Vol. 41, No. 3, Aug. 1992, pp. 243-
254.
[ 29] S. EL- Dolil, et . al. , Telet raffic Performance of Highway Microcells wit h Overlay Macrocell , I EEE Journ. of Select .
Areas in Comms. , Vol. 7, No. 1, Jan. 1989, pp. 71- 78.
[ 30] W. A. Massey, W. Whit t , The Highway PALM: A St ochast ic Model t o Capt ure Space and Time Dynamics in
Wireless Communicat ion Syst ems , Proc. I TC 14, 1994, pp. 503- 511.
[ 31] M. Frullone, P. Grazioso, A. M. Serra, "Performance Evaluat ion of a Personal Communicat ion Syst em in a Urban
Environment ", 12
t h
I EEE VTC Conf. , Denver, May 1992, pp. 498- 501.
[ 32] I . Seskar, et . al. , Rat e of Locat ion Area Updat es in Cellular Syst ems , Proc. I EEE Veh. Tech. Conf. , Denver, May
1992.
[ 33] Amot z, Bar- Noy and I lan Kessler, "Tracking Mobile Users in Wireless Communicat ions Net works", Proc.
I NFOCOM ' 93, 1993, pp. 1232- 1239.
[ 34] B. Awerbuch and D. Peleg, "Concurrent On- line Tracking of Mobile Users", Proc. ACM 9/ 91, 1991, pp. 221- 233.
[ 35] V. Anant haram, et . al. , "Opt imizat ion of a Signaling Hierarchy for Mobilit y Tracking in Personal Communicat ions
Net works", I nt er. Telet raffic Congress, I ndia, Nov. 1993, pp. 45- 51.
[ 36] S. S. Rappaport , " The Mult iple- Call Hand- Off Problem in High- Capacit y Cellular Communicat ions Syst ems", I EEE
Trans. on Vehic. Tech. , Vol. 40, No. 3, Aug. 1991.
[ 37] J. C. - I . Chuang, "Aut onomous Frequency Assignment and Access for TDMA Personal Port able Radio
Communicat ions", I EEE VTC ' 91, St . Louis, MO, May 19- 22, 1991.
[ 38] G. Riva, " Performance Analysis of an I mproved Dynamic Channel Allocat ion Scheme for Cellular Mobile Radio
Syst ems", 42
nd
I EEE VTC, Denver, May 1992.
[ 39] J. C. - I . Chuang, "Performance I ssues and Algorit hms for Dynamic Channel Assignment ", I EEE Journ. on Select .
Areas in Commun. , Aug. 1993, pp. 955- 963.
[ 40] T. J. Kahwa and N. D. Georganas, " A Hybrid- Channel Assignment Scheme in Large Scale Cellular St ruct ured
Mobile Communicat ion Syst ems", I EEE Trans. on. Comms. , Vol. 26, Apr. 1978.
[ 41] S. M. Elnoubi, R. Singh, and S. C. Gupt a, " A New Frequency Channel Assignment Algorit hm in High Capacit y
Mobile Communicat ion Syst ems", I EEE Trans. on Vehic. Tech. , Vol. 31, Aug. 1982.
Mobilit y Modeling in Third Generat ion Mobile Telecommunicat ion Syst ems
Page 27
[ 42] M. Frullone, et . al. , " On t he performance of Packet Reservat ion Mult iple Access wit h Fixed and Dynamic Channel
Allocat ion", I EEE Trans. on Vehic. Tech. , Vol. 42, No. 1, Feb. 1993, pp. 78- 86.
[ 43] N. D. Wilson, et . al. , "Packet CDMA Versus Dynamic TDMA for Mult iple- Access in an I nt egrat ed Voice/ Dat a PCN",
I EEE Journ. Select . Areas in Communic. , Vol. 11, No. 6, Aug. 1993.
[ 44] S. Tekinay, B. Jabbari, " A Measurement - Based Priorit izat ion Scheme for Handovers in Mobile Cellular Net works" ,
I EEE Journ. Select . Areas in Comms. , Vol. 10, No 8, Oct . 1992.
[ 45] R. Vij ayan, J. M. Holt zman, A Model for Analysing Handoff Algorit hms , I EEE Trans. on Vehic. Tech. , Vol. 42, No.
3, Aug. 1993, pp. 351- 356.
[ 46] K. G. Abakoukmin, Design of Transport at ion Syst ems , 1986, ( in Greek) .
[ 47] Woot on H. J. and Pick G. W. , " A model for t rips generat ed by House Holds" J. Transport Econ. and Policy 1( 2)
1967.
[ 48] S. E. Evans, " A relat ionship bet ween t he gravit y model for t rip dist ribut ion and t he t ransport at ion problem in
linear programming", Transp. Research Vol. 7, Pergamon Press 1973, pp. 39- 61.
[ 49] L. J. Leblanc, " An efficient approach t o solving t he road net work equilibrium t raffic assignment problem" ,
Transport at ion Research Vol. 9, Pergamon Press 1975, pp. 309- 318.
[ 50] N. M. Mit rou, G. L. Lyberopoulos and A. D. Panagopoulou, Voice and Dat a I nt egrat ion in t he Air- I nt erface of a
Microcellular Mobile Communicat ion Syst em , I EEE Trans. on Vehic. Tech. , Vol. 42, No. 1, Feb. 1993, pp. 1- 13.

You might also like