Computer Communications 31 (2008) 3752–3759

Contents lists available at ScienceDirect

Computer Communications

journal homepage: www.elsevier .com/locate /comcom

Intelligent network functionalities in wireless 4G networks: Integration schemeand simulation analysis

Noureddine Boudriga *, Mohammad S. Obaidat, Faouzi ZaraiCNAS, SUPCOM, University of Carthage, Technopark Elghazala, Ariana, TunisiaComputer Science Department, University of Monmouth, NJ, USA

a r t i c l e i n f o

Article history:Available online 15 May 2008

Keywords:Wireless 4G networkIntelligent networkBest network selectionLocation managementNetwork performanceSimulation analysis

0140-3664/$ - see front matter � 2008 Elsevier B.V. Adoi:10.1016/j.comcom.2008.05.003

* Corresponding author. Tel.: +216 98645073; fax:E-mail addresses: nab@supcom.rnu.tn (N. Boudri

(M.S. Obaidat).

a b s t r a c t

In future wireless and mobile environments it is likely that mobile stations will be able to choosebetween multiple access networks offering competing services. Wireless 4G network should offer a wideuse of information processing techniques, efficient manipulation of the network resources, and reuse ofnetwork functions. It should also allow the insertion of supplementary capacities making it easy to addservices. In this article, first we study the design of intelligent services in wireless fourth generation net-works. Then, we propose several intelligent network functionalities to determine the best network to useefficiently. Finally, we study the effect of the intelligent functions on the network performance using sim-ulation analysis.

� 2008 Elsevier B.V. All rights reserved.

1. Introduction

The evolution of wireless network technologies has led to dif-ferent generations of wireless cellular systems referred as nG(1G, 2G, 2.5G, 3G,. . .). Current wireless systems only provide lim-ited services. For instance, 2G, 2.5G users are asking for communi-cation services at the wire line quality (both voice, data,multimedia) when they are mobile. A very high-data rate isrequired to realize this and this data rate is well beyond the capa-bility of the third generation (3G) wireless systems. This is themotivation behind the increasing research thrust on defining anddesigning wireless fourth generation (4G) networks [1–3]. Thevision for 4G and beyond systems is towards unification of variousmobile and wireless networks. However, there is a fundamentaldifference between wireless cellular and wireless data networks,such as WLANs. The difference is that cellular systems are com-monly circuit-switched, meaning that for a certain call, a connec-tion establishment has to take place prior to the call. On thecontrary, wireless data networks are of packet-switched nature [4].

Fourth generation wireless networks will be a heterogeneousnetwork consisting of different access networks, which may over-lap with one another. In this environment, a mobile station isequipped with a mobile device containing multiple wireless inter-faces or a multi-mode interface. It will enhance and extend mobil-ity: anytime and anywhere accessibility, IP mobility, privacy andsecurity of communications, diversity of services while keeping

ll rights reserved.

+216 71856829.ga), obaidat@monmouth.edu

low cost. The wireless 4G networks are expected to include wire-less access, wireless mobile, wireless LANs PANs, and satellite net-works and to provide a wide range of services including high-speeddata and real-time multimedia to mobile users. The mobile user isexpected to be able to communicate through different wirelessnetworking architectures and to roam within these architectures.

The wireless 4G networks will envision flexible and adaptiveintegration of network technologies to enable mobile node toseamlessly roam between access networks [1,2]. These networksshould not only provide the commonly known Internet services,but also should transport the traditional voice service and otherreal-time applications. Eventually, they should allow more ad-vanced broadband multimedia services with varying QoSrequirements (e.g., preferred low delay, limited jitter, highthroughput, or high peak packet rate). Mobility management,including, handoff, best network selection, and locationmanagement, allow multimedia applications to get certainquality guarantee on bandwidth, jitter and delay for its packetsdelivery [5–14].

Wireless 4G networks will not only help improving existing ser-vices [5,6] but integrate intelligent algorithms for mobility man-agement, resource management, access control, routing, etc. Theywill offer a wide use of information processing techniques, an effi-cient manipulation of the network resources, and reuse of networkfunctions. They should allow the insertion of supplementarycapacities making it easy to add services.

In public telephone networks, the experience realized with theintegration of intelligent services has shown its advantages andsuccess. The situation is now taking place with wireless secondand third generation networks [6]. Complementary services make

N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759 3753

the management of the wireless 4G networks easier, and provideflexible access to multimedia applications. To the best of ourknowledge, very few works have dealt with wireless intelligentnetworks. The Third Generation Partenership Project 2 (3GPP2)[6] has started taking interest in intelligent network in wirelessnetworks. It has just focused on defining the wireless intelligentnetwork architecture in terms of functional entities.

Our contribution in this article is 2-fold. First, we define anarchitecture for heterogeneous Wireless 4G networks that makesthe composing access networks transparent to users and providesseamless services. Second, we propose the introduction of comple-mentary services and intelligent network functionalities, such asbest network selection, handoff and location management, in wire-less fourth generation, showing their effects and behaviors.

The remaining part of this article is organized as follows. Sec-tion 2 describes the wireless 4G, requirements and characteristics.Section 3 introduces some complementary services. Section 4 givesthe design of some services that we find useful for multiple ad-vanced applications. Section 5 describes and analyzes the perfor-mance results obtained from the simulation analysis. Finally, theconclusion is presented in Section 6.

2. Wireless fourth generation

The fourth generation wireless networks will envision flexibleand adaptive integration of network technologies to enable mobilenode to seamlessly roam between access networks [7,8]. Thesenetworks should not only provide the commonly known Internetservices, but also should transport the traditional voice serviceand other real-time applications. Eventually, they should allowmore advanced broadband multimedia services with varying Qual-ity of Service (QoS) requirements (e.g., preferred low handoff delay,high bandwidth utilization, or low packet loss rate). It also can usehigh-altitude platforms (HAPs) as they will have an important roleto play in future systems and applications. HAPs have the potentialto exploit many of the best aspects of terrestrial and satellite-basedsystems (LEO, GEO), while offering advantageous propagationcharacteristics. Such platforms may be airships or aircraft and forenvironmental considerations would ideally be solar powered[15,16].

2.1. Fourth generation requirements

The main objectives of 4G wireless networks can be stated ashaving ubiquity, and multi-service platform, with secured accessand traffic. Wireless 4G service quality will be the collective effectof the performance of all system elements in combination with theuser expectations, which determines the degree of satisfaction ofthe 4G users. The main requirements are shown below:

– Seamless access. Seamless access in wireless 4G network willmean connectivity to the end user across a wide range of heter-ogeneous access technologies and access networks using differ-ent technologies with minimal involvement from the user.

– Low handoff delay and loss rate. Handoff introduces packet lossand delay which can severely damage data communications.Handoff mechanisms must therefore be managed to minimizethese aspects and maintain a good network performance (nodisruption to user traffic, minimal additional signaling, andlow packet loss rate).

– Multi-service network. A multi-service network is an essentialproperty of the new wireless generation, not only because it isthe main reason for user transition, but also because it will givetelecommunication operators access to new levels of traffic.Voice will loose its weight in the overall user bill with the raise

of more and more data services. The wireless 4G network willoffer unlimited mobility and support high-data rate, serviceswith variable bandwidths, symmetrical and asymmetrical datatransfer (e.g., voice, video, fax, Internet services). This broadarray of services will be provided to mobile users by supportingload balancing, priorities and guaranteed quality of serviceclasses.

– Broadband wireless access networks. 4G network need to inte-grate means of delivering multimedia communications tometropolitan areas using High-Altitude Platform (HAP) systemsor/and satellite systems. In many ways, HAPs are equivalent tosatellite systems except that they are much cheaper and theycan easily be redeployed and/or maintained.

– Security access and traffic. To attain the success of wireless 4Gnetwork, these systems must address security issues properlyand integrate strong cryptographic schemes into the system.

2.2. Model of wireless fourth generation

The basic goals of the wireless 4G networks are to make the het-erogeneous network transparent to users and to design a systemarchitecture that is independent of the wireless access technology.There are several architectures using multiple different radioaccess networks (RANs) [3,?]. The main models are tunneled net-works, hybrid networks, and heterogeneous networks. The distinc-tion between these models is in the layer on which the RANscommunicate.

2.2.1. Tunneled networksIn this model, a user has a service agreement with the operators

of several RANs independently. Based on a certain policy, the opti-mal network for the requested service is selected. The connectivitybetween networks is based on relatively high network layers of theInternet (i.e., transport or session layers). This will require no mod-ification to existing access networks.

2.2.2. Hybrid networksIn this model, we have a hybrid core that interfaces directly

between RANs and the IP backbone. In this model, RANs imple-ment the network layer and the layers below.

2.2.3. Heterogeneous networksIn this model, there is a common core network that deals with

all network functionality and operates as a single network. Differ-ent RANs handle only those tasks that are specifically related to acertain radio access technology. In general, wireless access radioincorporates the physical and the data link layers only. Communi-cation between RANs belonging to the same common core networkis based on lower network layers (link layer or network layer). Thisreduces the overhead, and improves the network performance.

In our work, we choose the hybrid model. In fact, the advanta-ges of this model include fewer duplicate functions and more ad-vanced services at the network or data link layer (e.g., it canprovide a better handoff between RANs). In the tunnel model, theconnectivity between networks increases the service latency.Moreover, all the networks have their own infrastructure (e.g., sig-naling, network discovery, handoff, etc.) This makes it very difficultfor existing network systems to cooperate efficiently. The proposednetwork architecture for wireless 4G network is illustrated in Fig. 1[9].

The basic architecture of the wireless fourth generation net-work is divided into fourth levels. Lower levels are comprised ofhigh bandwidth wireless cells that cover a relatively small areaconnection over a larger geographic area. Also, the networks of

WPAN networks

Cellular Station

WMAN-WWAN Hybrid Unit

WMAN-station

IP Access Router

WLAN-WMAN Hybrid Unit

WPAN-WLAN Hybrid Unit

WLAN station

WPAN station

WLAN networks

WMANnetworks

WWANnetworks

ASN Gateway AAA

WLAN AAA server

WPAN AAA server

MSC or GGSN HLR

WLANGateway

WPANGateway

Fig. 1. Wireless fourth generation architecture.

3754 N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759

lowest levels can be expected to play a complementary role to thenetworks of highest levels.

– The lowest level is comprised of a collection Wireless PersonalArea Networks (WPAN), which provide for lower power shortrange connectivity. Bluetooth provides the basis of this networklevel.

– The second level is a Wireless Local Area Network (WLAN). Itoffers mobile users high bandwidth wireless Internet connectiv-ity. The WPAN–WLAN Hybrid units provide the WPAN andWLAN protocol interoperability.

– The third level is comprised of a collection Wireless Metropoli-tan Area Networks (WMANs). The 802.16d and 802.16e providethe basis of this level. The WMAN–WWAN hybrid unit plays therole of the operation providing protocol interoperability.

– The final level is a Wireless Wide Area Network (cellular net-work), which provides a much lower bandwidth and covers amuch broader area than WMAN or WLAN. It provides connec-tion with the external network (IP backbone, etc.). The WirelessWide Area Network (WWAN) uses cellular networks for datatransmission and some examples of these cellular systems areGSM, GPRS, EDGE, CDMA 2000, WCDMA and UMTS.

To make this architecture complete, layer four can include theuse of high-altitude platforms and satellite-based access systems.

2.2.4. Inter technology hybrid unit (HU)The hybrid unit is a node serving as a bridge between two dif-

ferent access networks. It enables integration of various heteroge-neous networks. It plays the role of the operation providingprotocol interoperability and provides abstracted services to pairs

of different access technologies. The HU may support multiple basestations which provide wireless link-layer connectivity. The differ-ent types of hybrid units are:

– WWAN–WMAN hybrid unit. This unit is a node serving as a bridgebetween WWAN and WMAN networks. It can operate as a basestation for WWAN and subscriber station for WMAN.

– WMAN–WLAN hybrid unit. It has the ability to operate as a net-work access point for a WLAN with infrastructure network orcluster head for ad hoc network and subscriber station forWMAN.

– WLAN–WPAN hybrid unit. This unit can be used to interconnectWLAN and WPAN networks. It has the ability to operate as amaster for Bluetooth network and access point for WLAN net-work with infrastructure and cluster head for ad hoc network.

The Mobile Switching Center (MSC), the Serving GPRS SupportNode (SGSN), and the Home Location Register (HLR) are the maincomponents of cellular networks. The Access Service Network(ASN) gateway represents a boundary for functional interoperabil-ity with WMAN clients and WMAN connectivity service functions.The Authentication, Authorization and Accounting (AAA) server isa network server used for access control. Authentication identifiesthe user. Authorization implements policies that determine the re-sources and services that can be accessed by a valid user. Account-ing basically keeps track of time and data resources employed forbilling and analysis.

3. Intelligent network architecture

The experience with the public telephone networks has shownthat the integration of complementary services is not a simple taskfor at least two reasons. First, these services may need to be modifiedat any time, in a rapid and cheap manner. Second, complementaryservices need to be independent from network equipments.

The three major Intelligent Networks (IN) principles are serviceindependence, separation of basic switching functions from serviceand application functions, and independence of applications fromlower-level communication details. In order to support wirelessnetworks, mobility functions have to be added to these INprinciples.

The Service Switching Point (SSP) detects any request from asubscriber and communicates with the Service Control Point(SCP), which is the node containing the service execution logic,to obtain information on how to set up the connection. The SCPmay ask the SSP to send to the subscriber a recorded messageand can also collect a response from the subscriber and send it tothe SCP for processing. The Service Node (SN) provides service con-trol, service data, specialized resources and call control functions tosupport bearer related services.

For wireless 4G network, information on roamers is obtainedfrom that subscriber’s Home Location Register (HLR) or AAA server.Each subscriber is associated with a single HLR or AAA server,which retains the subscriber’s record. When a phone call goes toa subscriber’s MSC (or SGSN or wireless gateway) the MSC recog-nizes that the subscriber is roaming and asks the HLR for the sub-scriber’s location. The HLR will communicate that information tothe VLR and relay a temporary location number received fromthe visited system.

The SCP function may be located in the SSP node or it may be astandalone node. In the proposed architecture (as depicted in Fig. 2that we have built for the purpose), the SSP is typically located inthe wireless switcher (MSC, SGSN, or wireless gateway) and inthe hybrid unit. The HLR or AAA servers are usually a network ele-ment such as an SCP.

r

R

Fig. 3. Cell with WLAN integration.

N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759 3755

4. Intelligent hybrid unit changes

This category includes the dynamic busy hybrid unit changing,and economic power (or reduction of interference level) services.

4.1. Dynamic busy hybrid unit changing

It allows the access network (for each level) to find free connec-tions to hybrid unit when the required application is replicated.This service can be activated if it is impossible to connect the userto given destination hybrid unit.

4.2. Economic power services

The goal of this service is to reduce the interference level. Toexplain how the economic power service works, we give an exam-ple of integration of WLAN in 3G networks [9]. In a 3G network,terminals that are far away from the base station need to use anover proportional large amount of transmission power. In the relaycommunication case, i.e. when using an intermediate mobile or ac-cess point in between, the transmission power for the cellular linkcan be reduced. This fact has as a direct consequence in reducinginterference. Based on the observations above, an increase incapacity of the system is expected. The idea is illustrated inFig. 3. As shown in the figure, the 3G base station covers only theinner part of the cell with a radius r smaller than the nominal cellradius R. The area of the ring between this inner circle and the cellborder is covered by mobile relays using the ad hoc air interface (oraccess points). Since the capacity C of a cell depends on the area tobe covered we can define C as a function of the radius.

CðrÞ � CðRÞ with r � R

Hence, the relative capacity gain can be defined as:

g ¼ CðrÞ � CðRÞCðRÞ

The problem we face with this service is how to make SCPaware of the hybrid unit states (busy, out of order). The solutionwe propose is to duplicate SSP to all hybrid unit’s switches (MSC

Fig. 2. Intelligent wireless 4G architecture.

or GGSN for cellular network, ASN gateway for WMAN, WLANgateway for WLAN, WPAN gateway for WPAN). Any time a hybridunit returns a connection signaling message, SSP analyzes it andkeeps the SCP informed. The messages indicate the connectionacceptance, rejection, and termination.

To check the state of hybrid unit (active or idle), every time ahybrid unit’s switch sends a setup message to a hybrid unit, it trig-gers timer. If the timer expires Nmax successive times, then theswitch aborts the request and concludes a failure state. WhenSCP receives such information, it updates its database (SDP) whichcontains information such as hybrid unit identifier, and the num-ber and states of the available connections. SCP can then knowwhich hybrid unit to choose to satisfy the client request.

5. Intelligent best network selection services

This category includes the optimal choice of access technology.In fact, for wireless 4G, the first issue deals with how to be bestconnected. Given that a user may be offered connectivity frommore than one technology at any time, one has to consider howthe terminal and an overlay network choose the radio access tech-nology suitable for serving the user. For example, a WLAN is bestsuited for high-data rate indoor coverage. General Packet RadioService (GPRS) or Universal Mobile Telecommunications System(UMTS), on the other hand, are best suited for wide coverage andcan be regarded as wide area networks, providing a higher degreeof mobility. Thus a user of the mobile terminal or the networkneeds to make the optimal choice of radio access technologyamong all those available. After a network is allocated to a request-ing user, the QoS offered to that user must be monitored so thatQoS contracts are maintained. At a particular point during the life-time of the session, an intersystem handoff may be required. Ahandoff algorithm should both determine which network to con-nect to as well as when to perform a handoff between the differentnetworks. Ideally, the handoff algorithm would assure that the bestoverall wireless link is chosen.

To clarify this idea, we give an example of handoff between twohybrid units, as depicted in Fig. 4.

In the example, the Universal Mobile Station (UMS) is handing-off between two hybrid units (step1), from the old base station tothe new base station. Each BS is associated with an old HU and newHU. The steps of intelligent best network selection in handoff areshown below:

– Upon handoff to the new BS, the UMS first performs authen-tication with a new HU and configures a care-of-address andsends a message M1 describing both the source and destina-tion of the handoff (identities of the BSs and the IP address ofthe old HU).

Old HU New HU

UMS

(1)(2)

(3)

Old BS New BS

Fig. 4. Handoff inter-HU.

3756 N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759

– Then, the new HU sends M2 to the old HU. M2 contains the iden-tities of the two BSs (old and new), the identity of the UMS, theavailable bandwidth, and cost.

– Upon receipt of the message M2, the old HU checks the validityof the information in M2. If the latter is considered valid, the oldHU stores the tuple <new HU, new BS, bandwidth, cost>. Then, itreplies to the new HU with the result of validation.

– If old HU accepts M2, the new HU stores the tuple <old HU, oldBS, bandwidth, cost>.

– When one mobile in communication enters into the area of thenew BS, it issues to its current HU a message containing a list ofBSs that are currently reachable.

– Upon receiving a list of BSs, the current HU translates the list ofBSs included in the message to a set of candidate HU using themapping present in its cache. Then, it returns to the UMS the listof candidate HU.

– Then the UMS select the best network.

6. Intelligent location management

One of the challenging tasks in the next wireless generation isto efficiently maintain the location of mobiles stations that movefrom one region to another. Location management is the set offunctions executed to discover the current attachment point ofthe mobile station for call delivery. Location management coverstracking (updating) functionality and paging functionality. Loca-tion registration involves the mobile terminal periodically updat-ing the network about its new location. This allows the networkto keep a track of the mobile station. To correctly determine theUMS’s current location, an exchange of an UMS’s location informa-tion between itself and the network is needed. The different classi-fications of the location update and paging schemes are thefollowing. Location Update (LU) strategies can be broken down intotwo main groups.

Static schemes. Presently, most location management schemesare static, where location updates occur on either periodic intervals(time-based method) or upon every Location Areas (LA) change(movement-based method). For the movement-based method,the whole coverage area is divided up into LAs. A LA is a fixedset of cells. In these schemes the registration area is the LAs. Whena UMS moves and goes into a new LA, it sends a location updatemessage. However, static location areas incur great costs withthe ping-pong effect. When users repetitively move between two

or more location areas, updates are unnecessarily continuouslyperformed.

Dynamic schemes. Dynamic location management is an ad-vanced form of location management where the parameters oflocation management can be modified to best fit individual usersand conditions [10]. Theories have been and continually are beingproposed regarding dynamic location updates and location areas[11,12]. Additionally, many are re-examining paging and mobilityparameters based upon these developments. Many of these pro-posals in dynamic location management attempt to reduce compu-tational overhead, paging costs, and the required number oflocation updates. However, many of these proposals are exces-sively theoretical and complex, and are difficult to implement ona large scale.

6.1. Intelligent location update

To overcome these effects, we propose an intelligent and dy-namic location management for wireless 4G network, which inte-grate tightly many different access networks. This intelligentscheme is designed to take multiple parameters, such as velocityand mobility patterns, in order to determine the most efficientlocation management scheme, location areas and paging area.The steps of this method are shown below:

– Initially, on the first arrival of a service request from an UMS, theSSP must assign one of the possible location managementschemes. In fact, there are many alternatives: static paging canused with a time-based scheme, dynamic paging can used witha time-based scheme, and finally static paging can be used witha static location area-based scheme.

– Then, based on the velocity and mobility patterns, the SSP deter-mines the most efficient alternative. In such an example, havingknowledge of a user’s past movements combined with the user’scurrent speed and direction allows efficient predictive estima-tion when determining a possible future location for paging.Therefore, location updates may not need to be as frequent,thereby reducing the overall location management costs. In fact,predicting the location of a mobile during a period of time wouldreduce the frequency of executing special mechanisms for loca-tion and sending requests for this.

For dynamic location update, we consider a mesh cell configu-ration allowing us to evaluate the optimal size of the registrationarea in order to reduce location update and paging costs.

The mobile station is located in two location areas. Each loca-tion area consists of kxk cells, where k is evaluated for each useraccording to its mobility pattern and incoming call rate.

In the proposed scheme, the UMS can be registered in agroup of two LAs, n and n + 1. The mobile station stores thesetwo location area identifications in its local memory. If theUMS moves from LA(n+1) to LA(n+2), it leaves the group of locationareas where it is registered (the location area identification n + 2is not stored in its local memory), and it triggers a location up-date message. The new set of LAs where the UMS is registered isLA(n+1) and LA (n+2). The group of location areas where the MT isregistered has two or more location areas. The content of this setis dynamic, and two consecutive sets for the same UMS alwaysoverlap (they have at least one location area in common). Forthis reason, this scheme can be seen as a hybrid method: anintermediate option between a dynamic and a static scheme.Furthermore, if there are more than two location areas pergroup, when the UMS receives a new incoming call or triggersa location update message, the system evaluates it and sends anew set of location areas to be stored. In this set, the UMS iscentred as much as possible. The use of this location manage-

N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759 3757

ment technique provides a hysteresis effect that alleviates thesignalling overload due to the ‘‘ping-pong” effect.

6.2. Intelligent paging

In the intelligent paging, the registration area is divided up intoseveral paging areas. First, the UMS is paged in the paging areawhere it is most likely to be located. This paging area includesthe cell where the last location update was triggered or the lastincoming call arrived. If the UMS does not answer, it is paged inthe second paging area. The paging areas are normally disjointedsets and so on.

The UMS can move from a paging area to another while thesystem searches for it. In order to prevent the system failingto locate the UMS, the paging areas are not disjointed sets. Infact, the second paging area includes the first paging area, andso on.

To efficiently run this method for each UMS, the SSP storesthe number of successful paging requests in each cell, and thecell where the last contact with the network was registered.With these data, the probability of the UMS being in each cellcan be estimated and the registration area can be divided intopaging areas (determination of the radius k for each mobilestation).

Table 1Simulation parameters

Parameters Values

Simulation parametersSimulation time 20 min

UMTS parametersCell radius 300 mMax BS transmit power 43 dBmMin BS transmit power 33 dBmMUD 0.7Orthogonality factor 0.5

7. Performance evaluation

In this section, a simulation study for the effect of the intelligentfunctions on the network performance is described. We show howthe intelligent services behave and if there is a need to take intoconsideration their presence when dimensioning an intelligentwireless 4G networks.

The simulated system is assumed to be composed of one UMTSbase stations, eight WLAN access points; see Fig. 5.

7.1. Simulation assumptions and parameters

For UMTS, we consider a realistic model, in which the mobile ini-tially chooses a speed that is uniformly distributed over interval[0,10 km/h]. We also choose a direction for motion that is uniformlydistributed over [0�,360�]. We assume that the arrival rate of newcalls follows a Poisson process with parameter k1 for the voice ser-vice and k2 for the data service. The duration of a call is exponentiallydistributed.

For WLAN, nodes are assumed to move according to the randomwaypoint mobility (RWM) model [7]. The pause time is 20 s with auniformly random speed between 0 and 15 m/s.

Hybrid unit

Access point

Base station

Fig. 5. Simulated network.

The assumed propagation model that we use in this simulationis the ‘‘cost 231 indoor office model” without floor losses, as it isgiven by Chakravorty et al. [8]:

Li;jðdBÞ ¼ 37þ 30 logðdi;jÞ

where di,j is the distance between i and j. The path loss in WLAN isgiven by:

Li;jðdBÞ ¼ 40þ 35 logðdi;jÞ

Table 1 represents the general system parameters chosen forthe simulation analysis. Such parameters are often used in most re-lated simulation studies such as [3].

7.2. Dynamic busy hybrid unit changing

In this experiment, we evaluate the performance of the networkin terms of Satisfied Request Rate (SRR). We define SRR as shownbelow:

SRR ¼ number of satisfied requestsnumber of requests

The evaluation consists of comparing the SRR for a networkwith activation of Dynamic busy hybrid unit changing service andfor a network without activation of Dynamic busy hybrid unitchanging service. We investigate this in a network containingone base station and eight access points. Among the latter, two ac-cess points are selected as hybrid units. We vary the total numberof users. The results that can be deduced from Fig. 6 include:

– Satisfied requests rate decrease when the number of usersincreases. This is because, more users are admitted.

– The intelligent service improves the satisfied requests rate by3%. This can be explained by the fact that when the number ofusers increases, intelligent service can satisfy more user

Thermal noise power �105 dBmTadd �14Tdrop �18 dBReq. (No) in up (down), voice 6 (9) dBReq. (No) in up (down), data 2.5 (3.5) dBVoice rate in up (down) 16 (16) kbs/sdata rate in up (down) 144 (384) kbs/sCall rate, voice 2 calls/hCall rate, data 1.3 calls/hAverage call length, voice 90 sAverage call length, data 600 s

WLAN parametersWLAN range 30 mNumber of Nodes 50RF band ISM 2.4 GHzData rate 2 Mbits/sAP Transmit power 32 mWAP sensitivity �83 dBmThermal noise power �126 dBTransmit power 1 mW

Req

uest

s S

atis

fact

ion

Rat

e (%

)

0

0.2

0.4

0.6

0.8

1

20 60 100 140 180 220Number of users

with activation of intelligentservice

without activation of intelligentservice

Fig. 6. Satisfied requests rate.

0.5

1

1.5

2

Pack

ets

loss

rat

e (%

)

Number of users

Without activation of intelligentserviceWith activation of intelligentservice

0

20 60 100 140 180 220

Fig. 8. Packets loss rate.

3758 N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759

requests. But at a certain value of user’s number, intelligent ser-vice cannot find connections on hybrid units due to their heavyload.

The second experiment was performed to calculate the handoffblocking probably in the two cases with and without intelligentservice. The latter is defined as the ratio of the number of handoffmobile dropped due to the unavailability of resources to the totalhandoff users. Fig. 7 show the following facts:

– Handoff blocking probability increases when the number ofusers increases. This is because, more users are admitted.

– Intelligent service improves the handoff blocking probability.This is explained by the reduction of interference level (increas-ing of capacity).

– Handoff blocking probability gain increases with the increase ofnumber of users. This is explained by the increasing interferencelevel with the increase of the nodes number.

7.3. Seamless handoff services

The parameter we are interested to measure with this service isthe packets loss rate. Fig. 8 shows that the packet loss rate in-

0

0.5

1

1.5

2

20 60 100 140 180 220 Number of users

Han

doff

blo

ckin

g pr

obab

ility

(%

) without activationof intelligentservice

with activationof intelligentservice

Fig. 7. Handoff blocking probability.

creases with the increase of the number of nodes. This can be ex-plained by the increase of the waiting time passed by a packet ateach intermediate access points or hybrid units along its path.We notice also that the selection based on the proposed methodguarantees a packet loss rate smaller than the traditional selectionbased on the RSS.

In fact one can observe three different behaviors in the intervals[0,100], [100,1680], and [180,240]. When the number of user issmaller than 100 the packet loss rate increases a little and thetwo methods (with and without intelligent scheme) have similarperformances. In the second interval, we observe that the increaseof the loss rate is important for both schemes and our schemestarts performing better, reaching a percentage of 1.22%, comparedto 1.26. When the number user increase in the third interval theslope of the loss rate is reduced and our scheme get better perfor-mance reaching better results (about 5% for 200 users and 8% whenthe number is 240).

7.4. Intelligent location management

This sub-section is dedicated to describing the results in termsof performance and benefits of the proposal intelligent locationmanagement. These experiments were performed to evaluate theperformance of the intelligent location management scheme interms of signaling overhead, utilization rate and handoff blockingprobably.

Fig. 9 depicts the signaling overhead as a function of the num-ber of users. The signaling overhead is defined as the ratio of sig-naling messages and packet carrying payload. From this figure,we can make the following observations:

– The location update and paging signaling overhead increaseswith the number of a UMN. This can be explained by theincrease of registration requests.

– The activation of intelligent location management servicesreduces the location update signaling cost. It replaces stringsof characters with single codes. It allows to have a gain of about15% when the number of user is 240.

Fig. 10 shows the variation of the used bandwidth. This figureshows a good potential for network capacity increase based onthe activation of intelligent location management method. For in-stance, in the case of activation of intelligent location managementmethod, the maximum utilization bandwidth percentage could be82.1%, whereas this percentage falls down to around 80% in the

0

4

8

12

16

20

24

28

20 40 60 80 100120140160180200220240Number of users

Sign

allin

g ov

erhe

ad (

%)

without activation of intelligent service

with activation of intelligent service

Fig. 9. Signaling overhead.

0

10

20

30

40

50

60

70

80

90

100

20 40 60 80 100 120 140 160 180 200 220 240

Number of users

Util

izat

ion

(%)

with activation of intelligent service

without activation of intelligent service

Fig. 10. Utilization rate.

N. Boudriga et al. / Computer Communications 31 (2008) 3752–3759 3759

case of without activation of intelligent location managementmethod. The latter improves the utilization rate by 2.1%. This canbe explained by the fact that when the number of users increases,intelligent network can reduce the overhead.

Fig. 11 shows that the handoff blocking probability also in-creases with the average percentage link utilization. This is espe-

0

0.5

1

1.5

2

20 40 60 80 100120140160180200220240

Number of users

Han

doff

blo

ckin

g po

babi

lity

(%)

without activation of intelligent services

with activation of intelligent services

Fig. 11. Handoff blocking probability.

cially the case, when the capacity reaches 80%, the networksbecome saturated, more than 1.4% of the calls are dropped. Handoffblocking probability gain increases with the increase of number ofusers. This is explained by the increasing signaling location withthe increase of the nodes number.

8. Conclusions

In this article, we have proposed method to integrate intelli-gent services in wireless fourth generation and intelligent algo-rithms for the UMS to determine the best network to use andin a way to use resources efficiently. Then, we have studiedthe effect of these intelligent functionalities on the network per-formance using simulation analysis. The numerical results devel-oped in this article show that the intelligent services improvethe satisfied requests rate, the signaling overhead, the bandwidthutilization rate, the handoff blocking probability and the packetsloss rate.

Our paper did not address the impact of enhanced mobile ter-minals of the performance of 4G systems. Multimode terminals,for example, help achieving better provision of QoS and supportmultiple air interface technologies [17].

References

[1] V. Marques, R.L. Aguiar, C. Garcia, J.I. Moreno, C. Beaujean, E. Melin, M. Liebsch,An IP-based QoS architecture for 4G operator scenarios, IEEE WirelessCommunications 10 (3) (2003) 54–62.

[2] K. Arentz, M. Haardt, W. Konhauser, W. Mohr, The future of wirelesscommunications beyond the third generation, Computer Networks 37 (2001)83–92.

[3] V. Gazis, N. Housos, A. Alonistioti, L. Merakos, Generic system architecture for4G mobile communication, VTC pring 2003.

[4] P. Nicopolitidis, M.S. Obaidat, G. Papadimitriou, A.S. Pomportsis, WirelessNetworks, Wiley, 2003.

[5] N. Boudriga, O. Hassairi, M.S. Obaidat, Intelligent services integration in mobileATM networks, in: ACM Symposium on Applied Computing (SAC), ACM, SanAntonio, TX, 1998, pp. 91–97.

[6] 3GPP2 X.S0004-700-E, Version 1.0.0, Wireless Intelligent Networks, March2004.

[7] TSG RAN WG4 TR25.942, Version 4.2.0, Radio Frequency System Scenarios, July2002.

[8] R. Chakravorty, P. Vidales, K. Subramanian, I. Pratt, J. Crowcroft, Performanceissues with vertical handovers: experiences from GPRS cellular and WLAN hot-spots integration, in: Proceedings of the IEEE Pervasive Communications andComputing Conference, PerCom 2004, March 2004.

[9] F. Zarai, N. Boudriga, M.S. Obaidat, A Novel solution for integrating WLAN andWPAN using UMTS TDD, in: Proceedings of International Symposium onPerformance Evaluation of Computer and Telecommunications Systems,SPECTS 2005, July 2005, pp. 205–213.

[10] F. Zarai, N. Boudriga, M.S. Obaidat, Performance evaluation of a novel schemefor QoS provision in UTRA TDD, Computer Communications Journal 29 (8)(2006) 969–982.

[11] Y. Xiao, Optimal fractional movement-based scheme for PCS locationmanagement, IEEE Communications Letters 7 (2) (2003) 67–69.

[12] I.F. Akyildiz, W. Wang, A dynamic location management scheme for next-generation multitier PCS systems, IEEE Transactions on WirelessCommunications 1 (1) (2002) 178–189.

[13] M. Archan, R. Abhishek, K. Das Sajal, An information-theoretic framework foroptimal location tracking in multi-system 4G wireless networks, Infocom(2004).

[14] A. Bhattacharya, S.K. Das, LeZi-update: an information-theoretic approach forpersonal mobility tracking in PCS networks, ACM-Kluwer Wireless Networks(WINET) 8 (2) (2002) 121–137.

[15] J. Thornton, D. Grace, C. Spillard, T. Konefal, T.C. Tozer, Broadbandcommunications from a high-altitude platform: the European HeliNetprogramme, Electronics & Communication Engineering Journal 13 (3) (2001)138–144.

[16] C. Spillard, D. Grace, G.P. White, B. Gremont, T.C. Tozer, High altitude platformsat York, in: International Conference on Next Generation Teletraffic andWired/Wireless Advanced Networking (NEW2AN’04), 2–6 February 2004, St.Petersburg, Russia.

[17] S. McCann, W. Groting, A. Pandolfi, E. Hepworth, Next generationmultimode terminals, in: Proceedings of the 5th IEE InternationalConference on Mobile Communication Technologies, 2004, pp. 143–147.