Final report on traffic estimation and services characterisation.pdf

Contract Number: IST-2000-28088

Project Title: Models and Simulations for Network Planning and Control of UMTS

Project Acronym: MOMENTUM

n

Information Report Number: IST-TUL_WP1_DR_PUB_200_WL_05_D1.4 Date of Delivery: 27.05.2003 Report Title: Deliverable D1.4: Final report on traffic estimation and

services characterisation Editor: Lcio Ferreira (IST-TUL) Authors: Lcio Ferreira (IST-TUL), Luis M. Correia (IST-TUL),

David Xavier (IST-TUL), llen Vasconcelos (IST-TUL), Erik Fledderus (TNO).

Reviewers Carlos Caseiro (Telecel/Vodafone),

Erik Fledderus (TNO)

Abstract: This final report addresses the main results achieved in WP1 on procedures to generate mobility and traffic scenarios, to be used in the deployment of UMTS radio networks and on service characterisation. After discussing the challenges concerning the generation of multi-service traffic, a service set is chosen and described in detail, classified and characterised, and the users profiles are established. A traffic forecast of static users is built for the city of Lisbon, as an example, based on an operational environment with users spread over it generating calls according to certain services usage patterns. Key parameters, necessary data, and interdependencies among data are identified and described in detail. A mobility scenario is defined, characterised by different mobility types and a mobility model that controls the movement of users on a motion grid. Traffic demand scenarios are then defined for dynamic, static and short-term dynamic simulations, where in particular average load grids are presented. User generation is addressed as well. Key word list: UMTS, Scenarios, Services, Users Profile, Traffic Estimation, Mobility, Average Load, IST, Key Action IV, Action Line IV.4.1 Key Action: IV, Essential Technologies and Infrastructures Action line: IV.4.1, Simulation & Visualisation Confidentiality: MOMENTUM PUBLIC

IST-TUL_WP1_DR_PUB_200_WL_05_D1.4 MOMENTUM PUBLIC

2/63

Document History

Date Version Comment Editor 17.03. 2003

1 First version. Lcio Ferreira (IST-TUL)

11.04. 2003

2 Second version updated with Carlos Caseiro (Telecel/Vodafone) Erik Fledderus (TNO), Alexander Martin and Oliver Wengel (TUD), Andreas Eisenblaetter (Atesio) and Ranjit Perera (UB) review comments.

Lcio Ferreira (IST-TUL)

09.05. 2003

3 Third version with updated BHCA tables, population distribution and resulting BHCA and load grids, and updated with second review comments by Erik Fledderus (TNO).


18.05. 2003

4 Fourth version with the inclusion of dynamic load simulator results.


27.05. 2003

5 Final version ready to be delivered.



3/63

Contents

Contents....................................................................................................................3

List of Figures ..........................................................................................................5

List of Tables............................................................................................................6

List of Tables............................................................................................................6

List of Notation........................................................................................................7

1 Introduction......................................................................................................9

2 UMTS Demand ..............................................................................................11 2.1 Key Drivers and Barriers for UMTS Demand .............................11 2.2 Forecasting Demand ....................................................................12

3 Service Set.......................................................................................................14

4 Traffic Estimation..........................................................................................19 4.1 Initial Considerations ...................................................................19 4.2 User Profile ..................................................................................20 4.3 Operational Environment .............................................................22 4.4 Population distribution.................................................................24 4.5 Subscribers distributions.............................................................27 4.6 BHCA grids..................................................................................29

5 Mobility Scenario...........................................................................................33 5.1 Introduction..................................................................................33 5.2 Mobility Model ............................................................................34 5.3 Penetration of Mobility Types .....................................................37 5.4 Implementation of mobility .........................................................41

6 Traffic Scenarios for Simulation ..................................................................46 6.1 Average Load Grids .....................................................................46 6.2 Simulation Approaches ................................................................48 6.3 Generation of Users .....................................................................49

7 Conclusions.....................................................................................................52

A Bearer specifications......................................................................................54 A.1 Uplink bearers ..............................................................................54 A.2 Downlink bearers .........................................................................56

B Updated Transition Tables............................................................................59


4/63

References ..............................................................................................................62


5/63

List of Figures

Figure 2-1: Percentage of residential respondents willing to use each of the services [7]........................................................................................12

Figure 2-2: Percentage of business respondents willing to use each of the services [7]........................................................................................12

Figure 2-3: Example of UMTS segment market share evolution. ..........................13 Figure 3-1: Service set bit rate range and DL session volume................................15 Figure 3-2: Service set traffic flow characterisation during a session (time and

bit rate domain) [2]. ..........................................................................17 Figure 4-1: General process for the construction of a traffic scenario....................20 Figure 4-2: Lisbon land use data thematic map [22]. ..............................................23 Figure 4-3: Lisbon vector data thematic map [22]...................................................23 Figure 4-4: Lisbon Operational Environment.........................................................24 Figure 4-5: Lisbon population distribution during the day. ....................................25 Figure 4-6: Calculation of persons per pixel in the different vector

operational environment classes. ......................................................26 Figure 4-7: Lisbon UMTS penetration, per customer segment. .............................28 Figure 4-8: Lisbon UMTS subscribers, per customer segment. .............................29 Figure 4-9: Video-telephony BHCA grids [calls / hour / 400 m2 pixel]. ................30 Figure 4-10: Mass-Market/Speech-telephony BHCA grid [calls / hour / 400

m2 pixel]............................................................................................30 Figure 4-11: Service set BHCA grids. ....................................................................32 Figure 4-12: Location Based BHCA grid [calls/hour/400 m2 pixel], for an

equal range scale representation. ......................................................32 Figure 5-1: Mobility scenario, identifying different mobility types associated

to the operational environment classes. ............................................33 Figure 5-2: Velocity probability density function [27]. ..........................................35 Figure 5-3: Possible pixel transition directions.......................................................37 Figure 5-4: Mobility grids for the area of Lisbon under study. ..............................42 Figure 5-5: Conversion of vector to pixel data. ......................................................42 Figure 5-6: Pixel crossed by a street. ......................................................................43 Figure 5-7: Example of conversion from vector to raster format. ..........................43 Figure 6-1: Load Grids. ..........................................................................................47 Figure 6-2: Speech average load grids for different times of simulation,

considering a simulation step of 1 second. .......................................48 Figure 6-3: New BHCA grids, considering the restrictions of unavailable

services in certain operational environments....................................49 Figure 6-4: Generation process of active users.......................................................50 Figure A-1: The EbNo !!!! BLER relations for the uplinklink bearers....................55 Figure A-2: The EbNo !!!! BLER relations for the downlink bearers. ....................58


6/63

List of Tables

Table 2-1: Customer segmentation. ........................................................................13 Table 3-1: Service set parameters [2]. ....................................................................18 Table 4-1: Number of calls per day per customer segment. ...................................21 Table 4-2: Busy hour usage per segment. ...............................................................21 Table 4-3: Average number of calls in the busy hour (BHCA) per service and

customer segment subscriber. ...........................................................22 Table 4-4: Momentum operational environment classes. .......................................22 Table 4-5: Persons per pixel in vector operational environment classes (for a

grid of 20 m x 20 m pixel size resolution)........................................25 Table 4-6: Weights per non-vector operational environment class, to be

applied in estimated population per pixel in vector operational environment classes where population in that pixel is 0. .................26

Table 4-7: Operational environment share between customer segments (in %).....27 Table 4-8: UMTS subscribers penetration, per segment and for a specific

operator. ............................................................................................27 Table 5-1: Mobility types average velocity and velocity variation.........................36 Table 5-2: Mobility types PDF parameters.............................................................36 Table 5-3: Probability of changing direction values, for each mobility type..........36 Table 5-4: Mobility types penetration table per operational environment class. ....37 Table 5-5: Possible mobility types for each service. ..............................................38 Table 5-6: Available services per operational environment class...........................38 Table 5-7: Mobility type penetration table per operational environment class

for Speech-telephony, Location based, MMS and E-Mail services. ............................................................................................39

Table 5-8: Mobility type penetration table per operational environment class for Web browsing and File Download services................................40

Table 5-9: Mobility type penetration table per operational environment class for Video-telephony and Streaming multimedia services.................40

Table 5-10: Transition array reference table, combining the possible array of sides with the user entrance side to the pixel. ..................................44

Table 5-11: Pixel oriented direction probabilities, for all mobility types. ..............44 Table 5-12: Specific loss per service and mobility type. ........................................45 Table A-1: Characterisation of the uplink bearers ..................................................54 Table A-2: Characterisation of the downlink bearers .............................................56


7/63

List of Notation

3G 3rd Generation 3GPP 3rd Generation Partnership Project Asy Asymmetric B Back direction BER Bit Error Rate BHCA Busy Hour Call Attempt Bid Bi-directional CBD Central Business District COST European Co-Operation in the Field of Scientific and Technical

Research CRC Cyclic Redundancy Check CS Circuit Switched DL Downlink E East EDGE Enhanced Data rates for GSM Evolution ETSI European Telecommunications Standards Institute F Forward direction FER Frame Erasure Ratio / Frame Error Rate GIS Geographic Information Systems GPRS General Packet Radio Service GSM Global System for Mobile Communications HSCSD High Speed Circuit Switched Data IST Information Society Technologies ITU International Telecommunications Union MM Multimedia MMS Multimedia Messaging Service MOMENTUM

Models and Simulations for Network Planning and Control of UMTS

N North NRT Non-Real Time NTB Non-Time Based O-M One to Many parties O-O One to One party PDF Probability Density Function PS Packet Switched RT Real Time S South


8/63

SF Spreading Factor SOHO Small Office/Home Office Sym Symmetric T Turn direction TB Time Based UL Uplink UMTS Universal Mobile Telecommunications System Uni Unidirectional W West WAP Wireless Application Protocol WP Work package WP1 Work package 1 Traffic Estimation & Service Characterisation WP2 Work package 2 Traffic Modelling and Simulations for

Interference Estimation WP3 Work package 3 Dynamic Simulations for Radio Resource

Management WP4 Work package 4 Automatic Planning of Large-Scale Radio

Networks WP5 Work package 5 Assessment and Evaluation WWW World Wide Web XML eXtensible Markup Language


9/63

1 Introduction

UMTS is intended to be a system providing a multiple choice of services and applications to users, enabling the mixed use of voice, video and data, partly at the will of the user, and partly depending on the availability of the network. This makes a huge difference from existing 2nd generation cellular systems, e.g., GSM, which were never foreseen for this purpose, although they have recently started to provide services other than voice or simple messaging. This poses a real challenge to those involved in the design and dimensioning of UMTS networks, coming from the fact that there is no real data available that can be used for the estimation of the traffic offered to the system. The foreseen variety of services, the enormous set of possibilities of their use, combined with the lack of solid marketing information, makes the task of traffic estimation a very difficult and challenging one. MOMENTUM [1] is devoted to the study of UMTS radio network planning, presenting a complete approach to the challenge of producing a realistic estimate of a location-variant demand distribution for mobile users. This is essential to generate and optimise a realistic network configuration that satisfies this demand. Services are characterised, usage profiles are built, and traffic and mobility scenarios are generated to model the future demand in the most realistic way while keeping at the same time the necessary flexibility to incorporate future insights. An optimised radio network configuration is achieved with a developed automatic planning tool, using heuristic rules for faster evaluation. To evaluate the performance of the obtained configuration, a powerful newly developed dynamic real-time system-level simulator is used, taking most dynamic aspects of UMTS into account. For every-day planning purposes, a fast and simple snapshot simulator will also be tuned to fit the results of the dynamic simulator the best way possible. A library of UMTS scenarios will be built and published, with test cases to be used as a benchmark in the development of planning tools. MOMENTUM deals with the dimensioning of UMTS radio networks in an optimum way, taking into account the relationships between services demand, traffic capacity and network performance. Thus, it is of key importance to establish mobility and traffic demand scenarios as accurately as possible, so that results coming from developed and/or used simulators, and from developed optimisation algorithms, make sense and can be used to really conclude on them. This report presents the final report of the work developed in WP1 [2], which tries to answer to the following question: Which time- and location-variant service demand distribution for mobile users is to be expected? It presents an approach to the problem of demand estimation, by presenting a clear characterisation of the foreseen services, and establishing a procedure for estimation of realistic traffic demand scenarios, based on actual population data and its characteristics, together with various assumptions on the use of services and on market forecasts. Given the fact that much of the data is related to geographical aspects, e.g., population distribution and clutter, a Geographic Information System (GIS) tool (MapInfo [3]) is used to visualise this information. MapBasic and C programming languages were combined to process data. A machinery to generate traffic scenarios is presented,


10/63

and all processing steps are identified. Key parameters controlling the generation of these scenarios, necessary data, and interdependencies among the various parameters are identified and described in detail. The procedure is a general one, i.e., it can be applied to any geographical area. Mobility scenarios are also addressed. They specify completely the motion of pedestrian users and vehicular ones on roads and streets. This allows the realistic simulation of a scenario with moving users generating traffic. Mobility will have an impact in the spread of the average load over the scenario. As an example of the needed data and processing for generation of these scenarios, the centre of Lisbon is illustrated. Besides this chapter, this document encompasses six others: Chapter 2 regards the demand of UMTS. In Chapter 3 the chosen UMTS service set used in the project is described. Chapter 4 is dedicated to traffic estimation generation. Chapter 5 addresses the mobility scenarios. In Chapter 6 the different traffic scenarios for simulation are presented. Conclusions are drawn in Chapter 7.


11/63

2 UMTS Demand

2.1 Key Drivers and Barriers for UMTS Demand

The standardisation of UMTS started already in the early 90s; the motto was to design a system that could deliver multimedia services anytime, anywhere. Especially during the late 90s, when the mobile industry boomed, the call for UMTS was strong: it would solve capacity problems, and it would bring the vision of ubiquitous computing close by. We know that since then, the Information and Communications Technologies [4] world changed considerably; after a deep dive down, most operators started to realise that customers are not willing to pay for just mobile Internet. It is clearer than ever that rolling out new wireless systems should go hand in hand with stimulating demand by actively investing in mobile data services. In this new era where realism and caution are the keywords, UMTS must try to regain its position. A key number of drivers and barriers [5], [6], [7] strongly influence the demand of UMTS: The realisation of the new technical possibilities of UMTS (high data rates,

symmetrical and asymmetrical connection, circuit- and packet-switched mode, support of simultaneous calls, etc) is of paramount importance for the success of this system.

The extreme fast development of fixed multimedia is a good indicator to assess the demand for UMTS. Nevertheless, the high cost gap between fixed and mobile may discourage the uptake.

The fast development of e-commerce is expected to have a good impact on the demand of UMTS.

With the increase in peoples mobility, nomadic workers appear as key UMTS customers, willing to pay for a continuity of broadband services outside the office while on the move.

The operators battle for UMTS customers deals with pricing, subsidies for terminals, and interesting applications. For customers, this will have a positive impact.

Network technologies such as HSCSD, GPRS and EDGE and the arrival of services such as WAP and despite its teething troubles i-mode will educate future UMTS customers with regard to data communications, and at the same time will give operators time to change from a circuit to a packet world.

The multitude of UMTS standards and the various options that are left open result in a limited availability of terminals have a negative impact on the uptake of UMTS.

Regulatory aspects and standards have enabled economies of scale and large visibility. Nevertheless, a number of operators did large investments on spectrum and licenses, when possibly the market share will be insufficient for all operators to do business.


12/63

2.2 Forecasting Demand

There are a multitude of studies into the likely take up of new mobile data services, carried out by the UMTS Forum [8], by leading analyst companies [9], [10], and others commissioned by operators [11] and vendors [12], [13]. It does appear from much of the research that there is a strong possibility that take up will be stronger than many pundits think. As an example, results of an inquiry to residential and business persons on their will to use certain mobile services are presented, Figure 2-1 and Figure 2-2, [7]. With an eye on recent developments, these figures may be interpreted in a relative sense, that is, the actual use will be very much influenced by the tariffs for each service.

Figure 2-1: Percentage of residential respondents willing to use each of the services [7].

Figure 2-2: Percentage of business respondents willing to use each of the services [7].

The definition of customer segments to identify typical user profiles is important for characterisation of UMTS demand. Three customer segments within MOMENTUM are considered Business, SOHO (Small Office /Home Office) and Mass-Market users and described in Table 2-1. Residential or Mass-market groups and a Business group are sensible choices for user groups: They are easily identified in terms of work, age, and income. They have particular patterns of using mobile services that are different enough

to treat separately.


13/63

They have different take-up times and rates, i.e., Business users are easily associated with the early adopters while the Mass-market falls more in the followers category.

A group that falls in between Business and Mass-market is the SOHO user group, also known as the small and medium-sized businesses; the spatial locations usually provide enough information to pinpoint this user group.

Segment Description

Business Early adapters, with intensive and almost entirely professional use, primarily during office hours.

SOHO Followers, with both professional and private use, during the day and in the evening. Mass-market With low use, with flat traffic levels.

Table 2-1: Customer segmentation.

Based also on the evolution of the GSM market in European countries, it is assumed that UMTS will first attract the high-end mobiles customers, mainly professionals who will require wideband capabilities while away from the office. SOHO users and the Mass-market segment will also be drawn to UMTS, not only because of the new services, but essentially because voice will be migrated onto this system. When UMTS is introduced, the mobile market will be quite close to saturation, and UMTS subscriptions will mainly replace existing ones. An example of segment market share evolution is given in Figure 2-3, identifying the evolution of the usage share of UMTS among segments. Note that this does not represent the true figures used in the scenarios, but merely illustrates the type of information needed.

Figure 2-3: Example of UMTS segment market share evolution.


14/63

3 Service Set

UMTS offers the technical possibility to provide a broad set of services and applications with different characteristics and target users. Data transfer, video-telephony, and multiple applications for E-commerce are foreseen, among many others, for deployment within UMTS, which constitutes an absolute novelty in mobile communications. In deliverable D1.1 [14], various perspectives into services classification proposed in the literature by different bodies (ITU-T, 3GPP, ETSI, and UMTS Forum) are presented. The 3GPP approach is taken for future work. In deliverable D1.2 [15], a detailed description and characterisation of UMTS services and applications is presented. First, service classes and taxonomy of parameters used for characterisation are presented. 25 foreseen services and 54 applications are then identified and described into detail. All services are classified according to 3GPP classes, applications are associated to each service and characterised, for the identified parameters. Most people now agree that there will not be a service that will conquer the market. Some claim it will be a number of small killer applications, or that it will be personalisation of services that are tailored to individuals needs. A set of 8 services is proposed in MOMENTUM for simulation [15], as a killer cocktail: Speech-telephony: Traditional speech-telephony. Video-telephony: Communication for the transfer of voice and video between

two locations. Streaming Multimedia: Service that allows the visualisation of multimedia

documents on a streaming basis, e.g., video, music, or slide show. Web Browsing: This is an interactive exchange of data between a user and a

web server. It allows the access to web pages. This information may contain text, extensive graphics, video and audio sequences.

Location Based Service: Interactive service that enables users to find location-based information, such as the location of the nearest gas stations, hotels, restaurants, and so on.

Multimedia Messaging Service (MMS): A messaging service that allows the transfer of text, image and video.

E-mail: A process of sending messages in electronic form. These messages are usually in text form, but can also include images and video clips.

File Download: Download of a file from a database. This killer cocktail is heterogeneous enough to meet the foreseen demands of future UMTS customers and to translate in simulations the diversity of services and traffic patterns UMTS bears. Several considerations have been taken into account for the choice of this set of services. This set is quite representative in terms of the foreseen services by several fora [8], [9], [10], [11], [12], [13]; as shown, e.g., in the market evaluation study presented in Figure 2-1 and Figure 2-2.


15/63

The four 3GPP service classes, grouping services according to specific characteristics and performance requirements, are well represented in the service set as described next [14]: From the conversational class (characterised by symmetric and real-time

conversational pattern services, with low emphasis on signal quality), Speech-telephony and Video-telephony services are chosen. The bit rate and session volume strongly differs between these two services, being important to handle this diversity in simulations.

From the streaming class (characterised by real-time almost unidirectional data flow applications with low delay variation, which can be processed as a steady continuous stream) Streaming Multimedia is chosen. This service covers both audio and video streaming.

From the interactive class (characterised by request-response pattern services, highly asymmetric, with low round trip delay and high signal quality) Web Browsing and Location Based services are chosen. The average DL session volume differentiates these two services;

From the background class (non real-time asymmetric services, with high signal quality), File Download, E-Mail and MMS services are chosen. File Download is a bi-directional service but highly asymmetric, most of the traffic being DL. The remaining services are differentiated by their average bit rate and DL session volume.

Detailed characterisation of services is presented in Table 3-1 following the taxonomy of parameters proposed in [15]. In particular, these services are very dissimilar in terms of Downlink (DL) session volume and indicative bit rate range, as shown in Figure 3-1. The traffic flow also results very diverse, as illustrated in Figure 3-2. A description of the services and source models for simulation purposes, in the XML MOMENTUM format, is presented in deliverable D5.2 [16].

Dat

aR

ate

[kbp

s]

400

0

80

160

240

320

40

120

200

280

360

0.1 1 10 100 1000

Data Volume [kByte]

VideoTlphny

StreamMM

SpeechLocationbased

MMS

Email FileDwnld W W W

Conversational

Streaming

Interactive

Background

3GPP Classes:

Figure 3-1: Service set bit rate range and DL session volume.


16/63

The description of some characterisation parameters used in Table 3-1 is presented below: Information type: sound, video, text, data, still image. Intrinsic time dependency: time-based (TB, where data blocks must be

displayed consecutively at predetermined time instants), or non-time-based (NTB).

Delivery requirements: real-time (RT, for immediate consumption), or non-real-time (NRT, stored for later consumption).

Directionality of Connection: unidirectional (Uni), or bidirectional (Bid). Symmetry of Connection (for Bid connections): symmetric (Sym), or

asymmetric (Asy). Number of Parties: one-to-one (O-O), or one-to-many (O-M). Switching mode: Packet Switched (PS), or Circuit Switched (CS). Source model: Final description of source models will be found in D2.7 [17].

These models will give more precise values or a full stochastic for the following parameters: "#the source bit rate and the average bit rate "#DL session volume

The bearers that are used to transport the information; when more than one possibility exists, the probability that a certain bearer is chosen is indicated, based on an eduacated guess.

Average Duration: average duration and DL session volume are directly related by the DL average source bit rate.

Maximum transfer delay: This is the maximum time used to transmit information through the air interface and the UMTS network.

Burstiness: ratio between peak and average bit rates. Block Error Ratio (BLER) target. Other parameters related with the mobility type are presented in chapter 6.


17/63

Figure 3-2: Service set traffic flow characterisation during a session (time and bit rate domain) [2].


18/63

Bit rate [kbps] Bearer

1 BLER

Class Service

Info.Type

TB

/NT

B

RT

/NR

T

Uni/B

id

Sym/A

sy

Parties

CS/PS

Source Models

Source Bit rate

range [kbps] UL DL UL DL

DL session volume

[kB]

Average Duration

[s]

Max. Transf. Delay

[s]

Burst- iness

UL DL

Speech- telephony Sound TB RT Bid Sym O-O CS

Speech Telephony 4 - 25 12,2 12,2 Speech Speech 91,5

2 120 0,15 1 - 5 0.010 0.010

C

o

n

v

e

r

s

.

Video- telephony

Sound Video TB RT Bid Sym O-O CS

Video- telephony 32 - 384 100 100 CS64 CS64 1500 120 0,15 1 - 5 0.002 0.002

S

t

r

e

a

m

.

Stream. MM MM

TB/ NTB RT Bid Asy O-O PS

3 Stream.

MM 32 - 384 3 60 - PS128 (10%) PS64 (90%) 2250 300 10 1 0.000 0.002

Web- browsing MM TB RT Bid Asy O-O PS

Web- browsing < 2000 1 30 -

PS384 (1%) PS64 (90%) PS32 (9%)

1125 300 4/ page 1 - 20 0.010 0.010

I

n

t

e

r

a

c

t

v

e

Location Based MM

TB/ NTB RT Bid Asy O-O PS

Location Based < 64 1 10 -

PS128 (1%) PS64 (90%) PS32 (9%)

22,5 180 0,2 1 - 20 0.010 0.010

MMS MM TB NRT Uni4 Asy O-O PS MMS < 128 30 30 PS64 (90%) PS32 (10%)

PS128 (1%) PS64 (90%) PS32 (9%)

60 16,2 300 1 - 20 0.010 0.010

E-Mail Data NTB NRT Uni Asy O-O PS E-Mail < 128 30 30 PS64 (90%) PS32 (10%)

PS128 (1%) PS64 (90%) PS32 (9%)

10 2,4 45 1 0.010 0.010

B

a

c

k

g

r

o

u

n

d

File Dwnld. Data NTB NRT Bid Asy O-M PS File Dwnld. 64 - 400 1 60 - PS128 (1%) PS64 (90%) PS32 (9%)

1000 132 0,5 1 - 50 0.010 0.010

Table 3-1: Service set parameters [2].

1 All bearers are DCH; the corresponding EbNo ! BLER table is given in the Appendix. The percentages behind the bearers indicate the (guessed) probability that this service is mapped to this bearer during a long simulation. 2: For the calculation of the equivalent Speech-telephony call volume, an activity factor of 50% was considered. 3: Streaming Multimedia can also be CS for the case of video streaming. 4: MMS and E-Mail are unidirectional services, existing in a session as an UL or DL transmission, but never both. 5: For the E-Mail maximum transfer delay, a server access of 4 seconds was considered.


19/63

4 Traffic Estimation

4.1 Initial Considerations

Once it is known which services future UMTS subscribers can use (the service set described in the previous section), a closer look to the more specific question where is the demand for these services? is taken. Several approaches have been found in the literature addressing the estimation of traffic demand. In [18] a detailed description of the mapping mechanisms leading to a traffic and mobility characterisation is provided, for a given combination of UMTS environments/services/QoS requirements/systems. In [19] theoretical and practical aspects related to the dimensioning of hybrid traffic for 3G systems are discussed, combining user profiles and geographical distribution of users concepts. In [20], a method for the estimation and characterisation of the expected tele-traffic in mobile networks is presented, based on a geographical traffic model obeying the geographical and demographical factors for the demand for mobile communication services. In [5], the evaluation of UMTS demand is analysed, presenting usage hypotheses and scenarios that provide a basis for estimating the traffic load /km to be handled by third-generation mobile systems. In MOMENTUM a global approach is used, combining several aspects of the ones observed in the literature. These are referenced along the description of the current approach. It corresponds to a simple but efficient way of estimating traffic demand, based on the available data from operators for the scenarios to characterise. The estimation of UMTS services usage corresponds to observe the following reality: An operational environment with UMTS users spread over it generating calls according to specific services usage patterns. A complete picture of the processing is illustrated in Figure 4-1. The three key elements to build a traffic scenario are: An user profile, describing how a subscriber generates calls; An operational environment; Spatial distributions of segmented subscribers, based on a population

distribution. The way each of these elements is built in order to generate a traffic scenario is described in this section. Taking into account the guidelines that were defined, scenarios are dimensioned for a desired deployment: a reasonable or extreme/worst case scenario in terms of service usage, a forecast for a certain year, a specific service usage forecast (e.g., not including speech, which could be independently supported by GSM), etc.


20/63

The visualisation of data has been done by using MapInfo GIS tool, and all the processing described in this section was performed with programming tools developed in MapBasic and C programming languages. As an example of the needed data and processing for generation of a traffic demand scenario, the centre of Lisbon is illustrated, an area of 4 km x 4 km. The presented data has a 20 m x 20 m resolution. Under the scope of the MOMENTUM project, from real data, several traffic demand scenarios for cities of Portugal (Lisbon and Porto), Netherlands (The Hague and Bilthoven) and Germany (Berlin, Hanover, Karlsruhe) have been generated according to different forecasts. Berlin, Lisbon and The Hague scenarios will be available at the MOMENTUM site as public scenarios, for benchmark in the development of planning tools.

Customer Segments Op. Env. share [%]

Population distribution

Penetrationof UMTS

Subscribers

Subscribers grids

BHCA grids /service/segment

Daily Call Attempts

UMTS usage in the BH

BHCA table

BHCA grids /service

User Profile

Operational Environment

Traffic scenario

OperatorMarket Share

Customer Segments Op. Env. share [%]

Population distributionPopulation distribution

Penetrationof UMTS

Subscribers

Penetrationof UMTS

Subscribers

Subscribers grids

BHCA grids /service/segment

Daily Call Attempts

UMTS usage in the BH

BHCA table

BHCA grids /service

User Profile

Operational Environment

Traffic scenario



Figure 4-1: General process for the construction of a traffic scenario.

4.2 User Profile

To characterise the diversity of service usage patterns, three customer segments are considered Business, SOHO and Mass-Market users, as presented in Section 2.2. Each customer segment has a specific profile of usage, generating calls of each service according to a specific usage pattern. A table of service set usage is defined for each segment, characterising the call generation pattern for each service of the set. The used parameter to characterise each service usage by a user is the Busy Hour Call Attempt (BHCA), which indicates the average number of calls performed in the busy hour. In this way, the user profile is characterised by service set BHCA tables. BHCA tables are built based on marketing data. They are dependent on many factors such as the country under study, specific marketing strategy of pretended UMTS usage, etc. They can be adapted, e.g., to a general increase of services usage by subscribers of a certain customer segment, or an increased use of specific services. To build these tables, a similar approach to the one used in [5] is


21/63

followed. First the total number of calls per day a subscriber of each customer segment performs is estimated. In Table 4-1, an example for Lisbon is presented for the year 2005. First simulations using the BHCA-grids derived from the numbers for the common settings as defined in [2] shows that the offered traffic is by far higher then the traffic. Analysis of the basic assumptions revealed that the BHCA-assumptions included in Report No. 6 of the UMTS Forum [21] are much lower. Taking into account that the numbers in the UMTS Forum report are based on a market study from 1997 where the whole mobile market was much more optimistic than today, the UMTS Forum numbers can be seen as an upper limit. Telecel/Vodafone, taking into consideration that the figures should not exceed the UMTS Forum values, estimated the presented values for Lisbon.

Service Business SOHO Mass-

Market Speech-telephony 4.167 2.400 1.768 Video-telephony 0.900 0.864 0.679 Streaming multimedia 0.600 0.576 0.170 Web browsing 0.400 0.256 0.075 Location Based 0.023 0.022 0.013 MMS 0.230 0.221 0.078 E-Mail 0.138 0.110 0.087 File Download 0.180 0.115 0.068

Table 4-1: Number of calls per day per customer segment.

A busy hour usage per customer segment is also estimated [5], as being the percentage of traffic per day taking place during the busy hour. In Table 4-2, busy hour usage values per customer type are presented.

Customer Segment Busy hour usage [%] Business 20 SOHO 15 Mass-market 7

Table 4-2: Busy hour usage per segment.

It can be seen from the above tables that Business users use UMTS services mostly on specific (busy hours) times of the day, whereas the demand from the Mass-market is evenly spread. By multiplying Table 4-1 and Table 4-2, a BHCA table per user type can be built, as presented in Table 4-3. According to the prediction for the scenario, all values are specified, resulting in this final table. The three chosen customer segments represent early adapters (Business users), followers (SOHO users) and the Mass-market. By changing the penetration and usage of UMTS services in each group, we are able to assess e.g. an early, medium or mature market situation. The characteristics of each service (average duration, rate, etc) can be specific per customer segment, evidencing once more the flexibility of the machinery.


22/63

Service Business SOHO Mass-

Market Speech-telephony 0.833 0.360 0.124 Video-telephony 0.180 0.130 0.048 Streaming multimedia 0.120 0.086 0.012 Web browsing 0.080 0.038 0.005 Location Based 0.005 0.003 0.001 MMS 0.046 0.033 0.005 E-Mail 0.028 0.017 0.006 File Download 0.036 0.017 0.005

Table 4-3: Average number of calls in the busy hour (BHCA) per service and customer segment subscriber.

4.3 Operational Environment

To build a traffic scenario for a certain city, the identification of the different existing operational environment classes is essential. This characterisation, that is intended to be as realistic as possible, has to translate the diversity of the scenario, identifying regions with similar characteristics in terms of land use and usage. A set of classes to characterise the operational environment is proposed and characterised by MOMENTUM in Table 4-4. Class Description Water Sea and inland water (lakes, rivers). Railway Railway. Highway Highway. Highway with traffic jam

Traffic jam in a highway, corresponding to a lot of cars stopped, or moving at a very low speed.

Road Main road of relatively high-speed users, typically inserted in suburban and rural areas. Street Street of low-speed users, typically inserted in an urban area.

Rural

Rural area, with low building and high vegetation density; Area with low population density, mainly of residential and primary sector

population; Little commerce.

Sub-urban

Sub-urban area with medium building and vegetation densities; Area with medium population density, mainly of residential and secondary

sector population; Little commerce.

Open Small pedestrian land area (square, open area, park, large pedestrian areas along streets) surrounded by mean urban, dense urban, or residential areas.

Urban Area with high building density and low vegetation density; Area with high population density, mainly of tertiary sector with some

residential population. Central Business District (CBD)

Area with very high building density, very high buildings, with almost no vegetation.

Area with very high population density, with tertiary sector population.

Table 4-4: Momentum operational environment classes.


23/63

For each city, operators have provided a large set of data for the specification of an associated operational environment. For the Lisbon public scenario, the example being presented in this deliverable, Telecel/Vodafone MOMENTUM partner has provided a large set of data [22] consisting of: Raster land use data: a pixel grid of 20 20 m2 resolution with information of

Vodafone specific land use classes (water, buildings, open areas, etc.) of each pixel, presented in Figure 4-2;

Vector data: identifying streets (highways, main roads, streets), railways, and coastlines configurations, illustrated in Figure 4-3.

Figure 4-2: Lisbon land use data thematic map [22].

Figure 4-3: Lisbon vector data thematic map [22].

A mapping is made of the specific Vodafone/Telecel raster and vector classes onto the MOMENTUM operational environment classes [2]. The resulting operational environment grid for Lisbon is presented in Figure 4-4.


24/63

Figure 4-4: Lisbon Operational Environment.

4.4 Population distribution

The starting point to characterise the distribution of subscribers is the spatial distribution of population. To obtain a refined population distribution, some processing is needed. Available data for Lisbon consists of resident population and workplaces per district, as well as population pendulum movement values [23] statistics of the number of persons entering or leaving each day the city from surrounding districts. As a first step, the distribution should correspond to the period of the day under study. This is obtained weighting residential population data with workplaces data per district, combined with pendulum movements of population in and out the scenario under study during the day, as described in [2]. The obtained population distribution, in the resolution of district areas, is presented in Figure 4-5 b). The ranges of the presented picture are determined according to an algorithm [24] such that the difference between the data values and the average of the data values is minimized on a per range basis. This reduces error and enables to obtain a truer data representation, resulting in a more refined visualisation of the spatial characteristics of distribution of population.

A more realistic and refined population spreading over the geographic scenario is needed for a resolution similar than the operational environment (20 m x 20 m for Lisbon). Weighting is applied according to the operational environment classes [2], to account for the different relative probability that a user in a certain district will be located at each operational environment class. As an example, population of a certain district will be more concentrated in CBD areas than in forests.. Users are in this way be spread in a more refined way. For the city of Lisbon, Figure 4-5 c) presents the resulting day population distribution weighted by the operational environment classes. It must be clear that this processing results simply in a better distribution of population. The total population per district and globally in the entire area under study is kept constant.


25/63

a) Common legend b) Original population data. c) Processed refined population data. [persons/km2]

Figure 4-5: Lisbon population distribution during the day.

For certain areas, the estimation of population was not precise enough. With the presented approach, it is difficult to have an estimation of population on a highway crossing a rural area without population (typical situation in some of the received data). Population needs to be independently estimated in the following situations: In highway, highway with jam, road and street pixels without population; In all highway and road pixels crossing rural or open areas (even if there is

population data); In all railway pixels. First is estimated that each car contains in average 1.5 persons. Evenly distributed cars are assumed, with a certain average distance between the cars, depending on the type of vector environment. In this way can be calculated how many persons are present on average on each pixel (the resolution of the final data is per pixel). In Figure 4-6 is presented the empirical way how, for each vector operational environment class, the number of persons per pixel is estimated. The accepted values by all MOMENTUM partners are presented in Table 4-5. Considering that the values presented in Table 4-5 correspond to a vector overlapping a CBD pixel, for the other non-vector classes (rural, suburban, open, urban) specific weights are applied to the presented values, as presented in Table 4-6, resulting in a final number of persons per pixel.

Class Persons/Pixel Street 2.4 Road 3.4 Highway 2.4 Highway jam 7.2 Railway 0.6

Table 4-5: Persons per pixel in vector operational environment classes (for a grid of 20 m x 20 m pixel size resolution).


26/63

Class Weight Water 0.2 Rural 0.2 Suburban 0.4 Open 0.6 Urban 0.8 CBD 1.0

Table 4-6: Weights per non-vector operational environment class, to be applied in estimated population per pixel in vector operational environment classes where population in that pixel is 0.

For different pixel sizes, a factor is applied in order to adapt these values. As an example, for a grid of 10 m x 10 m pixel resolution, values are divided by 2, since vector data is considered having always the width of a pixel. For 5 m x 5 m pixels all final values are divided by 4.

Street

Road

25 m

35 m

Highway

Highwayjam

1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

Railway

50 m

1.5 pers 1.5 pers 1.5 pers






1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers/car

200 pers/train

10 m

50 m

6500 m

Street

Road

25 m

35 m

Highway

Highwayjam

1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

1.5 pers 1.5 pers

Railway

50 m







1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers

1.5 pers/car

200 pers/train

10 m

50 m

6500 m

Figure 4-6: Calculation of persons per pixel in the different vector operational environment classes.

In other cases, a better estimation of the population was obtained extrapolating from GSM speech traffic data. This happened e.g. in exposition areas, train stations, or certain areas where no accurate population was available. From the operators GSM speech traffic data in the busy hour the number of persons was estimated, considering a certain fixed traffic per person (25 mErl for the case of Vodafone) and a penetration of the GSM operator (35% penetration for Vodafone). The combination of all these processings result in a refined population distribution, representing a good basis for the construction of a subscribers distribution.


27/63

4.5 Subscribers distributions

To build subscribers distributions, population is split into three customer segments (Business, SOHO and Mass-Market). The way population is split is new, dependent on the operational environment and specific to the type of available data. In [19] e.g., the availability of residence and business demographical databases as well as road traffic databases allows a different approach for the construction of subscribers distributions. In MOMENTUM, population of each customer segment is spread differently over the operational environment, according to their characteristics (e.g., in CBD a percentage of Business users higher than Mass-market ones, and the opposite in a rural area). A customer segment share table is defined per operational environment class; this gives a spatial distribution of customer segments share according to each class, Table 4-7. Values were defined together with MOMENTUM operators, which have extended marketing sources and experience on these matters. For each customer segment, this table tells where customers spend their time during the period under consideration. This is an important characteristic of users, identifying the areas where they are typically present. Since users have a specific service usage, can already be foreseen that this effect will result in a specification of the localisation of usage of certain services.

Operational Environment Class

Business SOHO Mass- market

Water 35 35 30 Railway 20 40 40 Highway 60 30 10 Highway with traffic jam 60 30 10 Main road 30 40 30 Street 10 20 70 Rural 2 3 95 Sub-urban 5 15 80 Open 25 40 35 Urban 25 40 35 CBD 80 10 10

Table 4-7: Operational environment share between customer segments (in %).

Only a certain percentage of the total population in the scenario will be a UMTS subscriber of a certain operator. The penetrations of UMTS per customer segment and per operator market share estimate this percentage. For the example of Lisbon, penetration of UMTS is presented in Table 4-8 for the different segments, an operator market share of 45% being considered for 2005. These values are market and operator dependent, resulting from predictions how the market will evolve.

Business SOHO Mass- Market Penetration 11.25% 6.75% 2.25%

Table 4-8: UMTS subscribers penetration, per segment and for a specific operator.


28/63

The combination of Table 4-7, Table 4-8 and Figure 4-4 results in three UMTS penetration distributions, per customer segment, as presented in Figure 4-7. These pictures illustrate clearly the different penetrations of UMTS, depending on the operational environment class and the customer segment. Many effects resulting from the dimensioning of these tables are identifiable graphically, e.g.: Higher penetration of business UMTS subscribers in CBD (9.00%) than SOHO

(2.70%) or Mass-Market (2.13%) subscribers; Higher penetration of Mass-Market UMTS subscribers on streets (1.57%) than

business (1.12%) subscribers.

a) Business. b) SOHO. c) Mass Market. d) Legend [% persons].

Figure 4-7: Lisbon UMTS penetration, per customer segment.

Applying these penetrations to the refined population distribution results in three customer segments subscriber spatial distributions, illustrated in Figure 4-8. It is interesting to discuss some visual effects on the resulting segmented subscribers distributions (grids): The effect of the different non-uniform UMTS penetration distributions on the

population distribution results in a graphical distortion of the population distribution, Figure 4-5 c). As an example, for the SOHO subscribers, it can be seen how different are Figure 4-5 and Figure 4-8 b), where in many areas the spatial distribution has increased/decreased relatively.

In CBD areas crossed by streets, which can be identified in Figure 4-4, the number of Mass-Market subscribers is higher on streets than on CBD areas, Figure 4-8 c), as dimensioned in Table 4-7, even if the population grid, Figure 4-5, specifies the opposite (containing less people on streets than on CBD area). The opposite happens for the Business segment, Figure 4-8 a), where more users are present in CBD areas than on streets.

The higher Business subscriber density area does not happen on the higher population density area, an Urban area. It happens on a CBD area, where the effect of the operational environment share percentage (25% in Urban versus 80% in CBD) results in a higher subscribers density on the CBD area.

The effect of different UMTS penetration values per segment (11.25% for Business versus 2.25% for Mass-Market) results, almost in the entire scenario, on a higher Business users density than Mass-Market one.


29/63

a) Business. b) SOHO. c) Mass Market. d) Legend [persons/km2]

Figure 4-8: Lisbon UMTS subscribers, per customer segment.

It can be concluded that the combined effects of a refined population distribution, the operational environment and the UMTS penetration, result in a refined way of building segmented subscriber distributions, very dissimilar from the initial population distribution.

4.6 BHCA grids

The combination of the spatial distributions of subscribers with the BHCA table results in traffic forecasts for the services usage per customer segment. These are expressed in terms of BHCA grids, where for each unit of area (pixel), the average number of new calls in the busy hour is specified, per service and customer segment. 24 BHCA grids make this resulting traffic demand scenario, one per customer segment and per service, as illustrated in Figure 4-1. In Figure 4-9 a, b and c, the Video-telephony BHCA grids for Business, SOHO and Mass-Market users are presented, using natural break ranges, specific for the image of each segment. Values of BHCA are presented per pixel (in the case of Lisbon corresponding to a 20 x 20 m2 pixel).

a) Business. b) SOHO.


30/63

c) Mass-Market.

Figure 4-9: Video-telephony BHCA grids [calls / hour / 400 m2 pixel].

Comparing the usage of Video-telephony by Business segment versus Mass-Market segment, it can be observed that the effect of a higher UMTS penetration (11.25% vs 2.25%, from Table 4-8) and of a higher usage of Video-Telephony (0.180 call/h vs 0.048 call/h, from Table 4-3) of Business users than of Mass-Market ones, results globally on higher BHCA values for Business users than Mass-Market one (maximum values of 0.575 versus 0.043 call/h/pixel). Many other effects are directly related with the ones identified on the segmented subscriber distributions. These refined and assorted figures result from the high number of different available screws to create a rich traffic forecast, which can be adapted to an expected or desired reality. Each one of these BHCA grids is directly proportional to the corresponding segmented population grids, illustrated in Figure 4-8 (nevertheless, the different range system used custom ranges versus natural break ranges doesnt allow the direct comparison). This was expected, since the processing to obtain the BHCA grids corresponds to multiply the each customer segment subscriber grid by the corresponding factor obtained from Table 4-3, for each service and corresponding customer segment. If we compare the Mass-Market BHCA grid for Video-telephony and for Speech-telephony, Figure 4-9 a) and Figure 4-10, we observe that the resulting BHCA values are directly proportional to each other.

Figure 4-10: Mass-Market/Speech-telephony BHCA grid [calls / hour / 400 m2 pixel]

For each service, three BHCA exist, one per customer segment, allowing specific characteristics of the same service. As an example, a Business user speech call


31/63

might have average call duration of 3 minutes, while it can be dimensioned for 2 minutes for the remaining segments. This diversity can also be applied to quality parameters, bearers or priorities. Nevertheless, in MOMENTUM, services characteristics are considered similar among customer segments. In this way, for each service, the three BHCA grids can be added. This results in 8 BHCA, one per service. In Figure 4-1 the global processing to obtain these final grids is illustrated. In Figure 4-11 the resulting BHCA grids per service for Lisbon are represented. Common ranges allow the direct comparison of BHCA values between services. It can be seen how different the resulting service BHCA distributions are. Location based service is the one with lower usage. In fact, this service is the one having the lowest BHCA values in Table 4-3, for all segments. Speech-telephony, Web browsing and E-Mail are services with high usage, but with very different resulting distributions. Even knowing that for a specific segment, all BHCA spatial distributions are directly proportional, (e.g. Figure 4-9 and Figure 4-10), note that the resulting BHCAs per service are all different. This is due to the fact that each BHCA/service figure results from the combination of three uniquely weighted BHCA/service/segment figures. This evidences the importance of splitting in segments the calculation of BHCA grids, before being added.

a) Speech-telephony. b) Video-telephony. c) Streaming MM

d) Web Browsing. e) Location-based. f) E-Mail.


32/63

g) MMS. h) File Download. i) Common legend [calls/hour/400 m2 pixel].

Figure 4-11: Service set BHCA grids. Business users, relatively to SOHO and Mass-market users, are very strongly present in all distributions due to the high UMTS penetration Table 4-8 and high service usage Table 4-3. This results, for almost all services, in high BHCA values where business users predominate (e.g. CBD areas). Nevertheless, if for a certain service, the combination of the BHCA value and UMTS penetration would be higher for the Mass-Market segment than for Business one, the resulting BHCA service distribution would have high BHCA values, e.g., on streets, something that does not happen for any service. The Location based BHCA distribution, Figure 4-11 e), seems to be uniform from the presented picture. Nevertheless, this is an erroneous conclusion due to the common used scaling, Figure 4-11 i). If the same data is represented using an equal range scaling, as illustrated in Figure 4-12, it can be seen that the distribution is in fact very diverse in space.

Figure 4-12: Location Based BHCA grid [calls/hour/400 m2 pixel], for an equal range scale representation.

All these effects are achieved thanks to a high number of available intuitive screws, having a natural link with real and measurable data/parameters. This allows, for a specific area, the dimensioning of services traffic forecast distributions according to an expected or desired set of characteristics.


33/63

5 Mobility Scenario

5.1 Introduction

Mobility is one of the major characteristics of wireless systems. With the large range of services UMTS will support, the old anywhere, anytime wireless systems premise can be extended with UMTS to anywhere, anytime, anything, within a certain range. It represents a big challenge for cellular planning. In most of the environments, the mobility characteristics of the terminals have a direct influence on the cell radius, and in turn on the investment cost of the network. A recent investigation [25] has quantified this effect, and has shown that the investment cost can increase by as much as 60% in environments where high terminal speeds prevail. Based on these facts, it is important to characterise the diversity of mobility types existing in the operational environments, so that inherent mobility characteristics of each environment are properly taken into account in simulations, Figure 5-1.

Rural

Sub-urban

Urban

Water

Railway

Highway

Major streetMajor road

100% Pedestrian100% Vehicular/Highway

50% Static50% Pedestrian

10% Static30% Pedestrian60% Vehicular/Major Road

10% Pedestrian90% Vehicular/Major Street

Rural

Sub-urban

Urban

Water

Railway

Highway

Major streetMajor road

100% Pedestrian100% Vehicular/Highway

50% Static50% Pedestrian

10% Static30% Pedestrian60% Vehicular/Major Road

10% Pedestrian90% Vehicular/Major Street

Figure 5-1: Mobility scenario, identifying different mobility types associated to the operational environment classes.

Mobility scenarios are built in MOMENTUM for the more realistic characterisation of scenario of users characterised by specific mobility patterns. This is of special interest for dynamic simulations where, during simulation, motion of users is simulated in the most realistic way. Also for static simulations this is important. Mobility will have also an impact on the spread of load of average load grids, used for generating snapshots. Mobility scenarios will be characterised by:


34/63

A mobility type penetration table per operational environment class, for the generation of moving active users;

A mobility grid per mobility type, as an implementation that (a) constrains the movement of users with its mobility type to a certain area and (b) precisely describes user movement on sample time-level.

To generate a realistic and diversified mobility scenario, different mobility types are then identified, according to their speed and movement type: Static; Pedestrian; Vehicular (highway, main road, street and railway). For each mobility type, PDFs for speed and discrete direction of motion are presented using the proposed generic Momentum mobility model. These mobility types are then mapped onto the operational environment in certain percentages, Figure 5-1. When a user is generated in a certain pixel, a mobility type is randomly going to be attributed (and will in general remain fixed).

5.2 Mobility Model

Several sources have suggested mobility models, according to different criteria, pointing out the key parameters for model customisation. An overview of the main existing mobility models and key parameters for model customisation was presented in [2] describing the following models: Random Walk Modelling [26]; Mobility Model with Triangular Velocity Distribution [27]; Simulation of a Mobile Highway Traffic [28]; Mobility Models described in ETSI [29] for:

"#Indoor Office Scenario; "#Outdoor to Indoor and Pedestrian Scenario; "#Vehicular Scenario; "#Mixed-cell Pedestrian/Vehicular Scenario.

Mobility Model Described in COST 259 [30]; A model for simulation of mobility in MOMENTUM was proposed and described. It combines the Mobility Model with Triangular Velocity Distribution [27] (for velocity estimation) and the COST 259 mobility model [30] (for discrete direction of motion estimation). These models where chosen due to their simplicity, still accounting for the main mobility characteristics. In addition, considering that users move in a pixel grid mobility scenario, the resulting vector describing the probability of taking a direction is converted into a vector describing the probability of crossing a side to a neighbouring pixel by including the effect of speed, pixel size, sample time and holding time.


35/63

For the velocity PDF, the Mobility Model with Triangular Velocity Distribution [27] is used with a specific average and variation for each scenario (mobility type). Figure 5-2 represents the triangular distribution and the respective parameters. For the Direction of Motion Estimation, the mobility model described in COST 259 [30] was chosen. This probability is defined by (5.1) [30], where w/2, w-/2 and w are the weight factors corresponding to probabilities, and is the standard deviation of the direction distributions. Standard deviation is assumed to be equal for the four variables.

( )( ) ( ) ( ) ( )

++++

+++

=

+

+

2

2

2

2

2

2

2

2

2

2

2222

2/

22

2/2

2/2/ 21

211

iiii

i

ewewewewe

wwwp i

(5.1)

A new term was added to the original equation to provide symmetry of the direction function around rad. Both weight factors and standard deviation will be specified for each scenario.

v [ms-1] Vmax Vav Vmin

2/(Vmax-Vmin)

f(v)

Figure 5-2: Velocity probability density function [27].

In order to generate a realistic and diversified mobility scenario, different mobility types are proposed for simulation, according to their type of motion and speed: Static; Pedestrian; Main Road/vehicular; Street/vehicular; Highway/vehicular; Highway traffic jam/vehicular;; Railway/vehicular.. For each mobility type, PDFs for speed and discrete direction of motion are presented using the proposed mobility model, modelling in this way the different mobility patterns. For the average velocity and velocity variation, some values where taken from [27] and others defined together with MOMENTUM operators, which have a large experience on these matters. For the MOMENTUM chosen mobility types, Table 5-1 summarises these characteristics. Average velocity and


36/63

variation are equal, except for Highway vehicular mobility type, where cars never move below a minimum speed. The direction PDF is described by (5.1) with parameters presented in Table 5-2. The MOMENTUM mobility scenario is a pixel grid. Considering that user motion is limited to transitions between pixels, only four possible directions for the mobile unit, forward (0), back (180), left (90) and right (-90) are considered possible, as illustrated in Figure 5-3. From the PDFs of each mobility type, the corresponding direction probability value can be extracted for each of the four possible directions of motion. Direction probability values, for each mobility type, are presented in Table 5-3.

Mobility type Vav [ms-1] Vav [kmh-1] [ms-1] [kmh-1] Static 0 0.0 0 0.0 Pedestrian 1 3.6 1 3.6 Street/vehicular 10 36.0 10 36.0 Main Road/vehicular 15 54.0 15 54.0 Highway/vehicular6 22.5 81.0 12.5 40.5 Highway with jam/vehicular 1 3.6 1 3.6 Railway/vehicular 22.5 81.0 22.5 81.0

Table 5-1: Mobility types average velocity and velocity variation.

Mobility type w/2 w-/2 w Static - - - -

Pedestrian 5/8 5/8 1/4 /8 Street/vehicular 1/2 1/2 0 /8

Main Road/vehicular 3/14 3/14 0 /8

Highway/vehicular 1/8 1/8 0 /8

Highway with jam/vehicular 1/8 1/8 0 /8 Railway/vehicular 1/8 1/8 0 /8

Table 5-2: Mobility types PDF parameters.

Mobility type 0 90 180 Static 0 0 0 Pedestrian 40 25 10 Street/vehicular 50 25 0 Main Road/vehicular 70 15 0 Highway/vehicular 80 10 0 Highway with jam/vehicular 80 10 0 Railway/vehicular 80 10 0

Table 5-3: Probability of changing direction values, for each mobility type.

6 The highway/vehicular model for Germany has a Vav of 35 ms-1 (126 kmh-1); the velocity variation remains unaltered.


37/63

0

-90

180 IN

90

Figure 5-3: Possible pixel transition directions.

The mobility model is used in the dynamic simulators to generate movement. More precisely, the movement of users is sampled, i.e., at different time-instances the position of the active user is updated. The level of detail is defined by the granularity of the raster, i.e., the pixel size. When these aspects are combined, we can express the probability of changing position (or pixel) by the generic direction vector (Table 5-3), the velocity (Table 5-1), the pixel size, the sample time and the holding time. This last quantity is assumed to be memory less, or exponentially distributed.

5.3 Penetration of Mobility Types

For the generation of moving users, mobility types are mapped onto the operational environment classes in a more or less empirical approach. In Table 5-4, for each operational environment, the percentages of users generated within a certain mobility type are presented. The presented values are a rough estimate and were defined together with MOMENTUM operators, which have large experience and sensibility for these matters. Nevertheless, values can be changed, expressing once more the flexibility of all the defined machinery.

Mobility type [% of users] Operational Environment

class Static Pedestrian Street/

veh. Main road/

veh. Highway/

veh. Highway jam/ veh.

Railway/ veh.

Water

Railway 100 Highway traffic jam 100

Highway 100

Main road 5 95

Street 5 5 90

Open 10 90

Rural 10 90

Sub-urban 20 80

Urban 30 70

CBD 50 50

Table 5-4: Mobility types penetration table per operational environment class.

When a user is generated in a certain pixel, a mobility type is randomly allocated according to these percentages. As an example, in a main road environment, 5% of the generated users are pedestrian, while 95% are Main road/vehicular.


38/63

Available services are strongly dependent on the mobility type. Some services (e.g. video-telephony) are not supported when the user moves with high speeds (highway). High bit rates and high delay sensitivity of certain services restrict the possible mobility types of active users. In Table 5-5, the possible mobility types for each service are presented. Can be seen, e.g., that a street/vehicular user driving at 36 km\h average speed cannot use Video-telephony service.

Mobility type

Service Static Ped. Street/ veh. Main

road/ veh. Highway/

veh. Highway jam/

veh. Railway/

veh. Speech $ $ $ $ $ $ $ Video-tlphny $ $ $ Str. MM $ $ $ Web brow. $ $ $ $ Loc based $ $ $ $ $ $ $ MMS $ $ $ $ $ $ $ E-Mail $ $ $ $ $ $ $ File Dwnld $ $ $ # $

Table 5-5: Possible mobility types for each service.

In this way, mobility types associated to certain operational environment classes inhibit the availability of certain services. In Table 5-6, the available services per operational environment are presented. This table is obtained by combining Table 5-4 and Table 5-5.

Operational Environment Classes

Service

Water

Rail-w

ay

Highw

ay traffic jam

Highw

ay

Main R

oad

Street

Open

Rural

Sub-urban

Urban

CB

D

Speech $ $ $ $ $ $ $ $ $ $ Video-telephony # $ $ $ $ $ $ $ $ Str. MM $ $ $ $ $ $ $ $ Web browsing $ $ $ $ $ $ $ $ Location based $ $ $ $ $ $ $ $ $ $ MMS $ $ $ $ $ $ $ $ $ $ E-Mail $ $ $ $ $ $ $ $ $ $ File Download $ $ $ $ $ $ $ $

Table 5-6: Available services per operational environment class.

The impact of Table 5-6 can be introduced directly on the BHCA grids, where the non-available services in certain pixels will have a corresponding BHCA of 0 for that service. Some considerations are presented in what follows. Video-telephony is a service that is not available for the Main Road/Vehicular mobility type (see Table 5-5). Nevertheless, in the Main Road environment the possible mobility types are Main


39/63

Road/Vehicular and Pedestrian. In this way, this service is available on a Main Road environment (see Table 5-6), but only with a Pedestrian mobility type. Considering a Street operational environment, from Table 5-6 can be seen that in all services are available. Nevertheless, for the case of Streaming Multimedia, this service is not compatible with the Street/Vehicular mobility type, being available for this service only two mobility types: Static or Pedestrian. The corresponding mobility type penetrations from Table 5-4 have then to be rebalanced since Street/Vehicular mobility type is not allowed. The values will be updated in order to keep 100% total sum; one will then have 50% of probability that the Streaming Multimedia user will have a Pedestrian mobility type and 50% for the Static one. Taking into consideration the results of Table 5-6, for each service, a rebalanced Table 5-4 is generated, as presented in Table 5-7 to Table 5-9. These tables can be directly used in simulations to randomly associate a mobility type to a new generated service, in a pixel of a certain operational environment class.



veh. Main road/

veh. Highway/


Railway/ veh.

Water

Railway 100

Highway traffic jam 100

Highway 100

Main road 5 95

Street 5 5 90

Open 10 90

Rural 10 90

Sub-urban 20 80

Urban 30 70

CBD 50 50

Table 5-7: Mobility type penetration table per operational environment class for Speech-telephony, Location based, MMS and E-Mail services.


40/63



veh. Main road/

veh. Highway/


Railway/ veh.

Water

Railway


Highway

Main road 100

Street 5 5 90

Open 10 90

Rural 10 90

Sub-urban 20 80

Urban 30 70

CBD 50 50

Table 5-8: Mobility type penetration table per operational environment class for Web browsing and File Download services.

Mobility type [% of users] Operational

Environment class Static Pedestrian

Street/ veh.

Main road/ veh.

Highway/ veh.

Highway jam/ veh.

Railway/ veh.

Water

Railway


Highway

Main road 100

Street 50 50

Open 10 90

Rural 10 90

Sub-urban 20 80

Urban 30 70

CBD 50 50

Table 5-9: Mobility type penetration table per operational environment class for Video-telephony and Streaming multimedia services.


41/63

5.4 Implementation of mobility

5.4.1 Introduction

For a complete characterisation of a mobility scenario, besides a mobility type penetration table per operational environment class and per service (Table 5-7-Table 5-9) for the generation of moving active users, Table 5-4, a detailed implementation of the mobility types is essential for the control of users movement. The result is a mobility grid where each pixel contains enough local information to set the user in motion. One mobility pixel grid per mobility type, Figure 5-4, will completely specify the motion of all users of that mobility type in the scenario. For each pixel, the possible transition sides are specified, with associated transition probabilities values. In this way, users of a certain mobility type are always kept in the specific mobility grid. In particular, motion of vehicular users will be vector oriented, driving along streets or railways. As an example, a moving Major Road/Vehicular user will always drive in major roads. The user may turn in crossings, according to a certain probability. Transition between mobility types (e.g., to a Street/Vehicular mobility type) may be allowed under special circumstances only between certain mobility types and in specific connecting points. Mobility types are characterised (besides their speed) by their movement type and corresponding mobility grid: Static users are non moving users (once generated in a certain pixel, they

remain always there); in this way, no mobility grid is associated to this type; Pedestrian users are walking users that move freely in all operational

environments except Water, Railway and Highway operational environment classes, as illustrated in Figure 5-4 d);

Vehicular users are driving users, being their motion restricted to their corresponding operational environment class (Highway, Major Road, Street or Railway); specific vehicular mobility grids, Figure 5-4 a) to c), specify the allowed motion pixels for Lisbon. The area under study has no highway.

a) Street mobility grid. b) Major Road mobility grid.


42/63

c) Railway mobility grid. d) Pedestrian mobility grid.

Figure 5-4: Mobility grids for the area of Lisbon under study.

5.4.2 Vector to pixel grid conversion

The implementation of the mobility types is done using pixel grids of a certain resolution with specific information for each pixel. The data underlying the operational environments are of raster and vector type, hence, a conversion of vector data onto raster data must be performed in order to extract vector information of the linking sides of pixels. As an example, the simulator must have enough information about the Railway path to move users along it; information has to be clear enough in order not to place suddenly a Railway/Vehicular user in an urban area of buildings or into water! This leads to the need of a format that maintains vector information in a grid of pixels. In Figure 5-5 the conversion of vector data into raster format is illustrated, for the main road operational environment class of a certain area of Lisbon. To maintain the vector information in a grid of pixels, the key issue is to keep two types of information in the pixel grid: Identification and properly labelling of the pixels crossed by vectors; Identification, for each pixel, of the linking sides, North, East, South or West

(N, E, S and W respectively); more precisely, for each pixel P, an array of sides SP will contain binary information of each side (N,E,S,W), indicating whether the appropriate side links (1) or not (0) to an other pixel.

a) Main road data in Vector format. b) Main road data in pixel grid format.

Figure 5-5: Conversion of vector to pixel data.


43/63

As an example, for the illustrated Main road pixel in Figure 5-6, the linking sides are N and S, and the corresponding array of sides for this pixel looks as (1,0,1,0).

N

S

EW

Figure 5-6: Pixel crossed by a street.

Considering that the simulator knows from which side a Main road/vehicular user has entered the pixel (e.g. N), this information is enough to know that this user will move to the pixel linked to side S.

5.4.3 Updating the Direction Transition Tables

The presented pixel oriented direction probability model, illustrated in Figure 5-3, will be influenced by the array of sides. As a simple illustration of the required update, consider the Street mobility grid in Figure 5-7 (b): For a Street/Vehicular user entering pixel K from side N(orth), with SK =

(1,0,0,1), it makes no sense the existence of the two (turning) possibilities of 25%. In fact, the direction probability will be 100% W(est) (considering 0% probability of going back);

For a Street/Vehicular user entering pixel A from side E(ast), and considering SA = (0,1,0,0), the direction probability distribution can only be 100% to the E(ast) side;

For a Street/Vehicular user entering pixel G from S(outh), and considering

GS!

= (0,1,1,1), the only possible sides are E(ast) or W(est), which in principle will be 50% for each side (since it is not possible to determine a principal direction). Nevertheless, if the user enters pixel G from E(ast), W(est) (straight direction) should have a higher probability than S(outh) (turning left).

a) Vector street. b) Street in raster format.

Figure 5-7: Example of conversion from vector to raster format.

In this way, for each mobility grid, a reference table of direction probabilities should be made for each of the possible configurations of the array of sides and for


44/63

each of the possible entering sides. First, a generic reference table is created, consisting of a possible array of sides and related entrance sides (Table 5-10). The following abbreviations are used: F forward direction; T turn direction; B back direction.

Entrance side Array of sides N E S W

(1,0,0,0) (F,0,0,0) (0,1,0,0) (0,F,0,0)

(0,0,1,0) (0,0,F,0)

(0,0,0,1) (0,0,0,F) (1,1,0,0) (B,F,0,0) (F,B,0,0)

(1,0,1,0) (B,0,F,0) (F,0,B,0)

(1,0,0,1) (B,0,0,F) (F,0,0,B) (0,1,1,0) (0,B,F,0) (0,F,B,0)

(0,1,0,1) (0,B,0,F) (0,F,0,B)

(0,0,1,1) (0,0,B,F) (0,0,F,B)

(1,1,1,0) (B,T,F,0) (T,B,T,0) (F,T,B,0) (1,1,0,1) (B,T,0,T) (T,B,0,F) (T,F,0,B)

(1,0,1,1) (B,0,F,T) (F,0,B,T) (T,0,T,B)

(0,1,1,1) (0,B,T,F) (0,T,B,T) (0,F,T,B)

(1,1,1,1) (B,T,F,T) (T,B,T,F) (F,T,B,T) (T,F,T,B)

Table 5-10: Transition array reference table, combining the possible array of sides with the user entrance side to the pixel.

As an example, a user entering a pixel with array of sides (1,1,0,1) from W side, will have, from Table 5-10, (T,F,0,B) as resulting transition array. This transition array identifies that: N is a turning side; E is a forward side; W is a back side.

An adaptation must be then made of the direction probabilities, summarised in Table 5-11. When certain directions are not possible, the probability for the respective side(s) is 0%, the remaining probabilities being rebalanced in order to obtain 100% again. Resulting tables are calculated for all mobility types, and presented in Appendix B.

Mobility Type Forward Turn left Turn right Back

Static - - - -

Pedestrian 40 25 25 10

Street/Vehicular 50 25 25 0

Main Road/Vehicular 70 15 15 0

Highway/Vehicular 80 10 10 0

High

Final report on traffic estimation and services characterisation.pdf

Documents