Network Design and Optimization for Quality of Services in Wireless Local Area Networks using Multi-Objective Approach

Similar documents
Demand-based Network Planning for Large Scale Wireless Local Area Networks

P. Bruschi: Project guidelines PSM Project guidelines.

Examination Mobile & Wireless Networking ( ) April 12,

Investigation and Simulation Model Results of High Density Wireless Power Harvesting and Transfer Method

Social-aware Dynamic Router Node Placement in Wireless Mesh Networks

Evaluation of Instantaneous Reliability Measures for a Gradual Deteriorating System

Direct Analysis of Wave Digital Network of Microstrip Structure with Step Discontinuities

Receiver-Initiated vs. Short-Preamble Burst MAC Approaches for Multi-channel Wireless Sensor Networks

Mobile Communications Chapter 3 : Media Access

The Influence of Propagation Environment in a Live GSM Network

ECMA st Edition / June Near Field Communication Wired Interface (NFC-WI)

When answering the following 25 questions, always remember that there is someone who has to grade them. So please use legible handwriting.

TELE4652 Mobile and Satellite Communications

Pointwise Image Operations

Lecture 11. Digital Transmission Fundamentals

Lecture 4. EITN Chapter 12, 13 Modulation and diversity. Antenna noise is usually given as a noise temperature!

Base Station Sleeping and Resource. Allocation in Renewable Energy Powered. Cellular Networks

Lecture September 6, 2011

The Comparisonal Analysis of the Concept of Rectangular and Hexagonal Pilot in OFDM

Signaling Cost Analysis for Handoff Decision Algorithms in Femtocell Networks

IEEE COMSOC MMTC Communications - Frontiers

Pulse Train Controlled PCCM Buck-Boost Converter Ming Qina, Fangfang Lib

Double Tangent Sampling Method for Sinusoidal Pulse Width Modulation

A Perspective on Radio Resource Management in B3G

Adaptive CQI adjustment with LTE higher-order sectorization

B-MAC Tunable MAC protocol for wireless networks

DS CDMA Scheme for WATM with Errors and Erasures Decoding

Performance Evaluation of a MAC Protocol for Radio over Fiber Wireless LAN operating in the 60-GHz band

Wrap Up. Fourier Transform Sampling, Modulation, Filtering Noise and the Digital Abstraction Binary signaling model and Shannon Capacity

Transmit Beamforming with Reduced Feedback Information in OFDM Based Wireless Systems

Transmit Power Minimization and Base Station Planning for High-Speed Trains with Multiple Moving Relays in OFDMA Systems

Foreign Fiber Image Segmentation Based on Maximum Entropy and Genetic Algorithm

Key Issue. 3. Media Access. Hidden and Exposed Terminals. Near and Far Terminals. FDD/FDMA General Scheme, Example GSM. Access Methods SDMA/FDMA/TDMA

A NEW DUAL-POLARIZED HORN ANTENNA EXCITED BY A GAP-FED SQUARE PATCH

4.5 Biasing in BJT Amplifier Circuits

QoE Driven Video Streaming in Cognitive Radio Networks: The Case of Single Channel Access

2600 Capitol Avenue Suite 200 Sacramento, CA phone fax

Radio Resource Management in Beyond 3G Systems

Demodulation Based Testing of Off Chip Driver Performance

An off-line multiprocessor real-time scheduling algorithm to reduce static energy consumption

Comparing image compression predictors using fractal dimension

Analysis ofthe Effects ofduty Cycle Constraints in Multiple-Input Converters for Photovoltaic Applications

EXPERIMENT #9 FIBER OPTIC COMMUNICATIONS LINK

Phase-Shifting Control of Double Pulse in Harmonic Elimination Wei Peng1, a*, Junhong Zhang1, Jianxin gao1, b, Guangyi Li1, c

On the Scalability of Ad Hoc Routing Protocols

Modeling and Prediction of the Wireless Vector Channel Encountered by Smart Antenna Systems

Will my next WLAN work at 1 Gbps?

ECMA-373. Near Field Communication Wired Interface (NFC-WI) 2 nd Edition / June Reference number ECMA-123:2009

Digital Communications - Overview

Multi-Objective Cross-Layer Optimization for Selection of Cooperative Path Pairs in Multihop Wireless Ad hoc Networks

A Flexible Contention Resolution Scheme for QoS Provisioning in Optical Burst Switching Networks

Notes on the Fourier Transform

2010 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media,

Quantifying Application Communication Reliability of Wireless Sensor Networks

THE economic forces that are driving the cellular industry

Knowledge Transfer in Semi-automatic Image Interpretation

On the Efficiency of Shaping Live Video Streams

Review Wireless Communications

Investigation of Novel Ultrasonic Positioning Method Installed in Sensor Network

Robot Control using Genetic Algorithms

Memorandum on Impulse Winding Tester

Starvation Mitigation Through Multi-Channel Coordination in CSMA Multi-hop Wireless Networks

Lab 3 Acceleration. What You Need To Know: Physics 211 Lab

Adaptive Antenna Array for Reliable OFDM Transmission

Ultrawideband Normalized Radar Cross Sections of Distributed Clutter

THE Cisco visual network index report predicts a drastic

Lecture #7: Discrete-time Signals and Sampling

Lecture 13: Capacity of Cellular Systems

Mohammed Ali Hussain et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 2 (3), 2011,

On Eliminating the Exposed Terminal Problem Using Signature Detection

Anti-Jamming Schedules for Wireless Data Broadcast Systems

Wideband characterization of the urban PCS channel

A new image security system based on cellular automata and chaotic systems

EE 330 Lecture 24. Amplification with Transistor Circuits Small Signal Modelling

Detecting Multi-Channel Wireless Microphone User Emulation Attacks in White Space with Noise

EE201 Circuit Theory I Fall

Using Box-Jenkins Models to Forecast Mobile Cellular Subscription

WIRELESS networks are growing rapidly as demand for

Dynamic Sounding for Multi-User MIMO in Wireless LANs

DECEMBER 15-18, 1997 ADELAIDE, SOUTH AUSTRALIA. Department of Mechanical Engineering, Hanyang University. Haeng Dang-Dong, Seong Dong-Ku, Seoul, KOREA

Performance Analysis of High-Rate Full-Diversity Space Time Frequency/Space Frequency Codes for Multiuser MIMO-OFDM

5 Spatial Relations on Lines

March 13, 2009 CHAPTER 3: PARTIAL DERIVATIVES AND DIFFERENTIATION

Fuzzy Inference Model for Learning from Experiences and Its Application to Robot Navigation

Chapter 2 Summary: Continuous-Wave Modulation. Belkacem Derras

HIGH THROUGHPUT EVALUATION OF SHA-1 IMPLEMENTATION USING UNFOLDING TRANSFORMATION

ECE-517 Reinforcement Learning in Artificial Intelligence

Variation Aware Cross-Talk Aggressor Alignment by Mixed Integer Linear Programming

Dimensions. Transmitter Receiver ø2.6. Electrical connection. Transmitter +UB 0 V. Emitter selection. = Light on = Dark on

Dimensions. Transmitter Receiver ø2.6. Electrical connection. Transmitter +UB 0 V. Emitter selection. = Light on = Dark on

Channel Estimation for Wired MIMO Communication Systems

Research Article The RSU Access Problem Based on Evolutionary Game Theory for VANET

On the Security of Angle of Arrival Estimation

(This lesson plan assumes the students are using an air-powered rocket as described in the Materials section.)

Chutima Prommak and Boriboon Deeka. Proceedings of the World Congress on Engineering 2007 Vol II WCE 2007, July 2-4, 2007, London, U.K.

Multiple Load-Source Integration in a Multilevel Modular Capacitor Clamped DC-DC Converter Featuring Fault Tolerant Capability

Dimensions. Model Number. Electrical connection emitter. Features. Electrical connection receiver. Product information. Indicators/operating means

Motion-blurred star image acquisition and restoration method based on the separable kernel Honglin Yuana, Fan Lib and Tao Yuc

Comparitive Analysis of Image Segmentation Techniques

The Design and Evaluation of a Wireless Sensor Network for Mine Safety Monitoring

Transcription:

Chuima Prommak and Naruemon Waanapongsakorn Nework Design and Opimizaion for Qualiy of Services in Wireless Local Area Neworks using Muli-Objecive Approach CHUTIMA PROMMAK, NARUEMON WATTANAPONGSAKORN * Deparmen of Telecommunicaion Engineering Suranaree Universiy of Technology Nakhon Rachasima, 30000 Deparmen of Compuer Engineering * King Mongku's Universiy of Technology Thonburi Bangkok, 10140 THAILAND cprommak@su.ac.h, naruemon@cpe.kmu.ac.h * Absrac: - A muli-objecive wireless local area nework (WLAN) design models have been developed o opimize he nework qualiy of services. The proposed model combines hree problems ogeher, including he opimal access poin placemen, he frequency channel assignmen and he power level assignmen. In addiion, i accouns for user populaion densiy in he service area, raffic demand characerisics and he physical srucure of he service area. The design model aims o deermine a nework configuraion ha opimizes he nework qualiy of services in erm of he radio signal coverage and he daa rae capaciy o serve expeced user raffic demand in he service area. Numerical resuls and sensiiviy analysis is performed o analyze he improvemen of he nework performance. We found ha when we incorporae he issue of he user daa rae capaciy in he design model, we can grealy improve he qualiy of service in erm of daa rae requiremen while slighly degrading he signal coverage availabiliy. I is observed ha as he weigh facor of he user daa rae capaciy objecive increases from 0 o 1, he user saisfacion level increases abou 40% while he signal coverage availabiliy decreases abou 10%. Key-Words: - Muli-objecive, Opimizaion, Nework design, Wireless Local Area Neworks 1 Inroducion Wih he coninued growh and he expansion of he infrasrucure-based Wireless Local Area Nework (WLAN) deploymens, efficien nework design mehods are required so ha he resuling WLANs can provide high Qualiy of Services (QoS). An infrasrucure nework employs an access poin (AP) for cenral conrol of he communicaion beween wireless users paricipaing in a Basic Service Se (BSS). A coverage area wihin which wireless users are free o move around and ye sill remain conneced o he AP is called a Basic Service Area (BSA) which covers an area ranging from 20 o 300 meers in radius depending on he ransmiing power level and he radio propagaion environmens [1]. For large service regions, a cellular archiecure wih muliple BSAs can be used in which he APs are inerconneced via a wired disribuion infrasrucure o form a single sysem called an Exended Service Se (ESS). Fig.1 illusraes an ESS where hree BSAs exis. Noe ha some of BSAs in he ESS can overlap. Recenly research effors using simulaion ools [2] and analyical models [3-7] have been carried ou o sudy performances and qualiy of services in WLANs. In his paper, we aim o solve he problem of laying ou BSAs o cover a arge region and achieve high qualiy of services. In paricular, we aim o deermine he opimal nework ESS BSA 1 AP1 BSA 3 W1 W4 AP3 W2 W5 W3 AP2 BSA 2 ESS = Exended Service Se AP = Access Poin W = Wireless user Basic Service Area (BSA) Communicaion link Fig.1 Infrasrucure-based WLANs ISSN: 1109-2742 827 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn configuraion (i.e. he locaion, frequency channel and power level of each AP) in order o maximize he nework qualiy of services in erm of he radio signal coverage and he daa rae capaciy o serve expeced user raffic demand in he service area. The issue on he qualiy of signal in he arge service areas and he concerns abou he user daa rae capaciy are wo imporan merics o be accouned in deermining he opimal nework configuraion. However, he majoriy of he published papers do no seek all hree parameers of he nework configuraion and he aenion is focused on eiher one of he design aspecs. Tradiional works focus on he AP placemen problems. The design focus mainly on he signal qualiy aspec, aiming o maximize Signal o Noise Raio (SNR) [8] or minimize pah loss [9,10]. Oher works [11,12] consider he frequency channel assignmen problems for WLANs. Ref [11] aims a maximizing he oal received signal srengh whereas Ref. [12] aims a maximizing he coverage availabiliy. Ref. [13] considers boh he AP placemen and he frequency assignmen problems by maximizing daa rae capaciy of he nework bu no considering he signal coverage aspec. In his paper we propose a muli-objecive WLAN design approach, opimizing boh he signal qualiy and he daa rae capaciy aspec o solve he AP placemen, frequency channel and power level assignmen problem. Moreover, he proposed model accouns for user populaion densiy in he service area, raffic demand characerisics and he physical srucure of he service area. The res of he paper is organized as follows. The nex secion describes he problem definiion of he WLAN design model and gives he mahemaical formulaion of he design model. Secion 3 gives numerical resuls and discussion. Secion 4 provides conclusions. 2 Problem Formulaion The ask of WLAN design is o place a given number of access poins (APs) in a service area ha may be locaed on a single floor or range across muliple floors. The APs may be configured wih differen power levels and frequency channels. The power level and frequency channel of an AP, ogeher wih he environmen specific pah loss and an anenna radiaion paern, deermines he region (called Basic Service Area (BSA)) in which he AP can suppor raffic demand o/from wireless users. We propose a problem formulaion for WLAN configuraion design seeking he opimal locaion, frequency channel and power level of each AP in a service area, in order o opimize he nework qualiy objecives described below. 2.1 Nework Qualiy Objecives 2.1.1 Radio Signal Coverage Objecive We consider signal qualiy in he proposed nework design model because he service availabiliy of he nework depends on availabiliy of he radio signal and he level of inerferences in he area. To achieve a paricular daa ransmission rae, wireless users mus be wihin a cerain range of he received signal srengh and he SIR hreshold. Thus, an imporan design objecive is o maximize he signal coverage availabiliy. We evaluae he signal coverage availabiliy by defining Signal-Tes-Poins (STP) where he received signal srengh and he SIR are assessed. To maximize he signal coverage availabiliy is o maximize he number of STPs of which he received signal srengh and he SIR level are greaer han he specified hreshold. 2.1.2 Daa Rae Capaciy Objecive As he user populaion grows and mulimedia applicaions requiring higher daa rae spread, he obainable user daa rae (hroughpu of each wireless users) becomes an essenial concern in designing WLANs [1]. According o capaciy analysis of he CSMA/CA proocol used in WLANs, he average user obainable daa rae can vary depending on he number of acive wireless users on he AP. As he number of wireless users wih acive daa ransfer connecions o a paricular AP increases, he effecive AP capaciy decreases. Thus, he locaion of APs should be a funcion of he densiy characerisics of he wireless users as well. Nework race sudies [14-18] repor ha average obainable user daa raes does no depend merely on he number of wireless users exising in he service area, bu also on he aciviy of users. Addiionally, raffic volume in he nework correlaes wih user behavior [14]. User behavior in urn correlaes o he ypes of locaions where users are siuaed and he major aciviies users ypically pursue in hose locaions [14-18]. The following secions discuss he incorporaion of informaion abou characerisics of WLAN usage and raffic paerns ino he design model. 2.2 Demand Node Represenaion The demand node concep used in faciliy locaion problems describes he geographic paern of demand for reail goods and services [23]. The concep was exended o wide-area wireless nework design o represen he disribuion of expeced nework raffic in a service area [19,20]. In ISSN: 1109-2742 828 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn designing WLANs a demand node represens an individual prospecive wireless user in a service area. The definiion allows a designer o describe precisely he poenial number of wireless users and heir locaions, in order o appropriaely place APs o accommodae expeced raffic demand. In WLANs, users communicae hrough APs using he CSMA/CA proocol in which users compee for channel access and share AP capaciy. Therefore, informaion abou he number of users is required o calculae an average user daa rae whereas he informaion abou user locaions is needed o approximaely assign users o an AP based on an accepable radio signal level. Nework race sudies characerized he usage of WLANs in various environmens such as on universiy campuses [15,17], in corporae office buildings [14], in academic building [18], and in a large audiorium [16]. Similariies exis in nework usage characerisics among differen nework environmens [14-18]. Traffic load a APs depends on users level of daa ransfer aciviy in addiion o he number of wireless users siuaed wihin he radio coverage area of APs. Nework race sudies show a correlaion beween users level of daa ransfer aciviy and locaions where users are presen [14-17]. In he proposed WLAN design model, a user aciviy level (α ) accouns for he correlaion beween nework usage characerisics and user locaions. α is he percenage of wireless users in a sub-area of ype who are simulaneously acive in daa ransfer hrough APs. Acive users paricipae in medium conenion o gain access o a communicaion channel and share AP capaciy. We define hree ypes of sub-areas: T = {1, 2, 3} where 1 denoes privae sub-areas, such as offices, 2 denoes public sub-areas for unscheduled aciviies, such as suden lounges, and 3 denoes public subareas for schedule-based aciviies, such as classrooms. The remaining user (1 α ) are idle users who, alhough siuaed in a sub-area of ype, do no generae daa ransfer aciviy over he nework a a paricular ime and herefore do no affec AP capaciy [16]. An average user daa rae requiremen in sub-area of ype (R ) imposes a desired link rae ha should be available o acive users in average. 2.3 Muli-Objecive Problem Formulaion The WLAN configuraion design problem is formulaed as a Muli-Objecive Problem (MOP), which combines wo measures of nework service qualiies: radio signal coverage and daa rae capaciy. MOP seeks an opimal nework configuraion, i.e. he opimal locaion, frequency channel and power level of each AP in a service area. Le A denoes a se of APs used in he service area, where n is he oal number of APs required. Le Ω j = {p j, f j,(x j, y j, z j )} denoe a se of decision variables which are parameers assigned o ap j for j = 1, 2,, n. p j denoes he power level assigned o ap j. f j denoes he frequency channel assigned o ap j, and (x j, y j, z j ) denoes he coordinae (x j, y j ) on floor z j where ap j is locaed. Le G denoes a se of signal es poins (STPs) represening locaions for esing he received signal srengh and he SIR level. Each STP g h refers o a coordinae in hree-dimensional space (x h, y h, z h ), where z h is he floor where g h is locaed. Le U denoes a se of demand nodes where index indicaes he ype of sub-area where demand node i is locaed. U U is a se of demand nodes in sub-area ype. The posiion of demand node i wihin he service area is denoed by (x i, y i, z i ), where (x i, y i ) is he coordinaes on floor z i where he demand node i is locaed. The user aciviy level (α ) and he average daa rae requiremen (R ) specify he nework usage characerisics of he demand node. The se of demand nodes ogeher wih he sub-area classificaion and parameers specifying nework usage characerisics (α and R ) are given as inpu o he design process. u Oher decision variables include u and g. is a user associaion binary variable ha equals 1 if demand node i U associaes o ap j A; 0 oherwise. g is a signal availabiliy binary variable ha equals 1 if STP h G can receive a signal from ap j A; 0 oherwise. Le P is he se of candidae power levels (discree values) for variable p j. F is he se of candidae frequency channels for variable f j and O is he se of candidae locaions for AP placemen. Parameers in he design process are classified ino saic and dynamic parameers. Saic parameers do no change during he design process because hey depend solely on sandard requiremens and he characerisics of user aciviy in service area. Saic parameers specifying he physical signal requiremens (e.g., he received signal srengh (P Rhreshold ) and he SIR level (SIR hreshold )), user profiles (e.g., he user aciviy level (α ) and he average user daa rae requiremen (R )), and he daa rae capaciy of AP (C). Dynamic parameers are recompued each ime a variable changes value during he design process. ISSN: 1109-2742 829 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn Dynamic variables include received signal srengh ( P ), inerference level ( Inf ), and average R obainable daa rae ( r ). The mahemaical model i of MOP for he WLAN design is wrien as follows: Objecives: 1) Maximize signal coverage area Maximize g h G j A f1 = (1) G f 1 measures he signal coverage availabiliy. I is he normalized number of STPs which he received signal srengh and he SIR level are greaer han he specified hreshold. 2) Maximize user saisfacion β u T j A i U Maximize f = (2) 2 β U ( ) T f 2 measures he user saisfacion level. I is he normalized number of users ha can obain he required daa rae. β is a relaive imporan weigh of user ype. I is defined as he raio of he required daa rae of user ype o he maximum bi R rae capaciy of he AP, β =. C Consrains: u j A u u 1, i U ( P R P Rhreshold ) 0 ( P Inf SIR ) 0 R i U, j A u g g ( r R ) 0 i hreshold, i U, j A, i U, j A ( P R P Rhreshold ) 0 ( P Inf SIR ) 0 R h G, j A hreshold,, h G, j A, (3) (4) (5) (6) (7) (8) u { 0,1}, i U, j A (9) g { 0,1}, h G, j A (10) Consrain (3) specifies ha each user can associae o a mos one AP. The decision variable u can be equal o one if he received signal srengh ha user i received from he ap j (P R in dbm) and he SIR level wih respec o he ap j (he received signal srengh ( P in dbm) less he inerference level (Inf in dbm)) mee he receiver sensiiviy hreshold (P Rhreshold ) and he SIR hreshold (SIR hreshold ) as specified by consrain (4) and (5), respecively. In addiion, when u is equal o one, consrain (6) mus be saisfied. I ensures ha he average daa rae available o wireless user i which is a ype user ( r ) is greaer han he specified user daa rae (R ). The 802.11 capaciy model and he user aciviy paern correlaed wih he ype of sub-areas where users locae are incorporaed in his consrain o esimae he average daa rae ha he acive wireless user can obain [21,22]. u is equal o zero oherwise. Consrains (7) and (8) assess he radio signal qualiy a he STP h, esing he received signal srengh and he SIR level. The decision variable g can be equal o one if he received signal srengh a he STP h ransmied from he ap j ( P ) and he SIR level wih respec o he ap j (i.e., P R Inf ) mee he received sensiiviy hreshold (P Rhreshold ) and he SIR hreshold (SIR hreshold ). Oherwise, g is equal o zero. Consrains (9) and (10) specify ha variable u and g are binary {0, 1} variables, respecively. 3 Numerical Resuls Numerical experimens were conduced on he service area in he building wih four floors. The building comprised of classrooms, offices, laboraories, suden lounges, and a library. The dimension of each floor is 33m 21m. The service area is divided ino grids of size 1m 1m as shown in fig.6. The grid poins specify he STPs. In fig.7-10, he symbol represens he demand nodes locaed in public areas for scheduled aciviies, he symbol represens he demand nodes locaed in public areas for unscheduled aciviies, and he symbol represens he demand nodes locaed in privae areas. User aciviy levels corresponding o each i R R ISSN: 1109-2742 830 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn sub-area ype are based on sudies showing ha users in privae sub-areas are he mos acive nework users, followed by users in he public areas for unscheduled aciviies and hen users of public areas for schedule-based aciviies [14-18]. Similarly, he average user daa raes are aken from observed nework usage characerisics [14-18]. Table 1 summarizes he nework usage characerisics. Table 2 summarizes he inpu parameers of he nework design problem. The design aims for 95% coverage availabiliy a he edge of AP coverage areas. In his case, a fading margin of 5.75 db is applied in he signal coverage calculaion. We applied he proposed MOP o he WLAN configuraion design for he four-sory building. A scalarizing funcion (11) (a weighed sum of he objecives) is applied o conver a muli-objecive problem o a single objecive problem. Max F = w 1 f 1 + w 2 f 2 (11) The paching algorihm [13] is applied o solve he scalarizing funcion. The maximum poin found is a paricular poin on he Pareo fron. For example, in fig.2, F i is a scalarizing funcion when using a weigh se i (w 1i, w 2i ). F i * is a single poin on a feasible region boundary where he line defined by he weighed sum F i is angen. F i * is a paricular poin on he Pareo fron ha is he maximum of F i. 3.1 Effecs of weigh facors Sensiiviy analysis is conduced o sudy effecs of weigh facors on he WLAN qualiy of service in erm of he signal coverage availabiliy and he user saisfacion level. In paricular, we generae an approximaed Pareo fron by running he program many imes using differen weigh ses. Each weigh se converges o differen maximum poin on he Pareo fron. The resuls ploed in fig. 3 demonsrae his behavior. The ploed is obained by running he paching algorihm o solve he nework design opimizaion five imes, using five differen weigh ses (Q = 5) in which he weigh values are spread equally as wrien in Eq. (12). We use seven APs in his experimen. The poins in fig. 3 are he maximum poins found wih each se of weighs. Two end of he fron are a (f 1 =67.6%, f 2 =34%) and (f 1 =58%, f 2 =75.4%). w 1q = (q-1)/(q-1), w 2q = 1- w 1q (12) where q = 1, 2,, Q, Q = he number of differen weigh ses In fig.3, We can observe ha as he w 2q increases from 0 o 1, he user saisfacion level increases abou 40% whereas he signal coverage availabiliy decreases abou 10%. We can see ha when we incorporae he issue of he daa rae capaciy in he design model, we can grealy improve he qualiy of service in erm of daa rae requiremen while slighly degrading he signal coverage availabiliy. f 2 Pareo fron F n * F i * F 1 * Feasible region F i = w 1i f 1 + w 2i f 2 f 1 Fig.2 Weighed sum of he objecives and he Pareo fron ISSN: 1109-2742 831 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn 3.2 Disribuion of he Pareo fron We conduc anoher se of experimens using differen values of Q o observe he disribuion of he Pareo fron. Fig. 4 presens he resuls obained by using seven values of Q (Q = 4, 5,, 10) o generae weigh ses. The poins found wih a se of weighs generaed by each Q are depiced wih a differen shape. I can be observed ha he poins spread ou more oward he middle and he lower righ corner of he fron. The upper lef corner of he fron is around (f 1 =67.6%, f 2 =34%) where he weigh se is (w 1q =1, w 2q =0). We can draw a similar observaion ha slighly increasing value of w 2q can improve he user saisfacion level grealy while slighly degrading he signal coverage availabiliy. For example, a he weigh se of (w 1q =0.87, w 2q =0.13), he user saisfacion level increases 22% whereas he signal coverage availabiliy reduces 3.6% (i.e., f 1 =64%, f 2 =56%). 3.3 Effecs of he number of APs The las se of experimens aims o sudy effecs of he number of APs used in he nework on he signal coverage availabiliy and he user saisfacion level. The resuls of using differen number of APs (4, 5,, 10) are ploed in fig.5. In his se of experimens, Q is equal o 7. The resuls show ha he user saisfacion level is proporional o he number of APs. The more APs used in he nework, he higher level of user saisfacion. Increasing he number of APs used in he nework improves he user saisfacion level more han improving he signal coverage availabiliy. I can be observed ha as he number of APs increases from 4 o 10, he user saisfacion level increases almos 20% whereas he signal coverage availabiliy increases abou 3%. The reason is ha he more APs used, he more capaciy he nework has for accommodaing user raffic demand. Since a limied number of channels exis in he available frequency specrum for an 802.11 WLAN, a muli-cell nework deploymen (using high number of APs) requires ha some channels are reused. Reuse of frequency channels in neighbouring cells can cause inerferences which affec he signal coverage availabiliy in he service area. 4 Conclusion This paper presens a novel mahemaical model for a WLAN configuraion design which is formulaed as a Muli-Objecive Problem ha combines wo % Signal coverage % Signal coverage % Signal coverage 70 68 66 64 62 60 58 56 54 52 50 30 40 50 60 70 80 % User saisfacion Q=5 Fig.3 Resuls wih differen weigh ses (Q=5) 70 68 66 64 62 60 58 56 54 52 50 70 65 60 55 50 45 40 35 30 Q=10 Q=9 Q=8 Q=7 Q=6 Q=5 Q=4 30 40 50 60 70 80 % User saisfacion Fig.4 Resuls wih differen weigh ses (Q=4,5,,10) 10APs 9APs 8APs 7APs 6APs 5APs 4APs 20 30 40 50 60 70 80 % User saisfacion Fig.5 Resuls of using differen number of APs ISSN: 1109-2742 832 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn measures of nework service qualiies: radio signal coverage and daa rae capaciy. A scalarizing funcion is applied o conver a muli-objecive problem o a single objecive problem. Sensiiviy analysis is conduced o sudy effecs of weigh facors on he WLAN qualiy of service in erm of he signal coverage availabiliy and he user saisfacion level. From numerical resuls we can conclude ha incorporaing he issue of he daa rae capaciy in he design model can grealy improve he qualiy of service in erm of daa rae requiremen while slighly degrading he signal coverage availabiliy. Acknowledgmen: This work was suppored by he Thailand Research Fund (Gran number MRG5080381). References: [1] A. Hills, Large-scale wireless LAN design, IEEE Communicaion Magazine, Vol.39, 2001, pp. 98-104. [2] B. Aslam, M. Akhlaq and S. A. Khan, 802.11 Disassociaion DoS aack Simulaion Using Verilog, WSEAS Transacions on Communicaions, Issue 3, Vol. 7, March 2008, pp. 198-206. [3] X. Chen, H. Zhai, X. Tian and Y. Fang, Supporing QoS in IEEE 802.11e Wireless LANs, IEEE Transacions on Wireless Communicaions, Vol. 5, No. 8, Augus 2006, pp. 2217-2227. [4] J. Liu and Z. Niu, A Dynamic Admission Conrol Scheme for QoS Supporing in IEEE 802.11e EDCA, Proc. of he IEEE Wireless Communicaions and Neworking Conference, 2007, pp. 3700-3705. [5] Y. Kuo, C. Lu, E.H. Wu, and G. Chen, An Admission Conrol Sraegy for Differeniaed Services in IEEE 802.11, Proc. of he IEEE Global Telecommunicaions Conference, Dec. 2003, Vol. 2, pp.707-712. [6] J.-S. Liu, Analysis of Burs Acknowledgemen Mechanisms for IEEE 802.11e WLANs over FadingWireless Channels, WSEAS Transacions on Communicaions, Issue 5, Vol. 7, May 2008, pp. 383-392. [7] J.-S. Liu, AchievingWeighed Fairness in IEEE 802.11-based WLANs: Models and Analysis, WSEAS Transacions on Communicaions, Issue 6, Vol. 7, June 2008, pp. 605-615. [8] E. Tangpraser and N. Waanapongsakorn, Muli-Objecive Design and Opimizaion of Oudoor Wireless Lan Sysem, Proc. of he 21 s Inernaional Conference on Circuis/Sysems, Compuers and Communicaions, 2006, pp. 625-628. [9] S. Kouhbor, J. Ugon, A. Kruger, and A. Ruginov, Opimal Placemen of Access Poin in WLAN Based on a New Algorihm, Proc. of he Inernaional Conference on Mobile Business, 2005. [10] M. Unbehaun and J. Zander, Infrasrucure Densiy and Frequency Reuse for User- Deployed Wireless LAN Sysems a 17 GHz in an office environmen, Proc. of he IEEE Inernaional Conference on Communicaions, 2001, pp. 2535-2539. [11] R. C. Rodrigues, G. R. Maeus, and A. A. F. Loureiro, On he Design and Capaciy Planning of a Wireless Local Area Nework, Proc. of he IEEE Conference on Nework Operaions and Managemen Symposium, 2000, pp. 335-348. [12] C. Prommak and B. Deeka, On he Analysis of he Opimal Frequency Planning for Wireless Local Area Neworks, Proc. of he 4 h PSU Engineering Conference, 2005, pp. 77-81. [13] X. Ling and K. Lawrence Yeung, Join Access Poin Placemen and Channel Assignmen for 802.11 Wireless LANs, Proc. of he IEEE Wireless Communicaions and Neworking Conference, 2005, pp. 1583-1588. [14] M. Balazinska and P. Casro, Characerizing mobiliy and nework usage in a corporae wireless local-area nework, Proc. of he Inernaional Conference on Mobile Sysems, Applicaions, and Services, San Francisco, CA, USA, May 2003. [15] D. Koz and K. Essien, Characerizing usage of a campus-wide wireless nework, Technical Repor TR2002-423, Deparmen of Compuer Science, Darmoun College, March 2002. [16] A. Balachandran, G. M. Voelker, P. Bahl, and P. V. Rangan, Characerizing user behavior and nework performance in a public wireless LAN, Proc. of he Inernaional Conference on Measuremen and Modeling of Compuer Sysems, June 2002. [17] D. Koz and K. Essien, Analysis of a campuswide wireless nework, he 8 h Inernaional Conference on Mobile Compuing and Neworking (MobiCom), Sepember 2002. [18] D. Tang and M. Baker, Analysis of a local-area wireless nework, he 6 h Inernaional Conference on Mobile Compuing and Neworking (MobiCom), Augus 2000. [19] K. Tuschku, Demand-based radio nework planning of cellular mobile communicaion ISSN: 1109-2742 833 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn sysems, Proc. of he IEEE INFOCOM, Vol. 3, 1998, pp. 1054-1061. [20] N. Weicker, G. Szabo, K. Weicker, and P. Widmayer, Evoluionary muliobjecive opimizaion for base saion ransmier placemen wih frequency assignmen, IEEE Transacion on Evoluionary Compuaion, Vol. 7, 2003, pp. 189-203. [21] F. Cali, M. Coni, and E. Gregori, Dynamic uning of he IEEE 802.11 proocol o achieve a heoreical hroughpu limi, IEEE/ACM Transacions on neworking, Vol. 8, 2000, pp. 785-799. [22] M. Heusse, F. Rousseau, G. Berger-Sabbael, and A. Duda, Performance Anomaly of 802.11b, Proc. of he IEEE INFOCOM, 2003. [23] Z. Drezner, Chaper 14, Faciliy locaion: A survey of applicaions and mehods, 1995, pp. 301-330. Table 1 Nework usage characerisics Sub-areas User aciviy level Average user daa rae (Kbps) Type 1: Privae sub-areas α 1 = 0.50 R 1 = 460 (such as graduae suden and library saff offices) Type 2: Public sub-areas for unscheduled aciviies α 2 = 0.40 R 2 = 260 (such as library sudy areas, suden lounge) Type 3: Public sub-areas for schedule-based aciviies (such as classrooms, laboraories) α 3 = 0.35 R 3 = 80 Table 2 Nework parameers used in he muli-objecive opimizaion for WLAN design Parameer Definiion Value Candidae se for Variables: P Se of candidae power levels for variable p j {0, 7, 13, 15, 17, 20, 24} in dbm F Se of candidae frequency channels for variable f j {2.412, 2.437, 2.462} in GHz Saic Parameers: α User acive level defines percenage of wireless users in sub-area ype ha are engaged in daa ransfer aciviies (i.e., paricipaing in channel conenion and See Table 1 sharing AP capaciy) R Average user daa rae requiremen in sub-area ype P Rhreshold Received sensiiviy hreshold -80 dbm SIR hreshold Signal o inerference raio hreshold 10 db C Daa rae capaciy of he ap j for j A 11 Kbps Pah loss Parameers: d 0 Reference disance d 0 1 meer n Pah loss exponen 3.3 δ Sandard deviaion represening shadow fading 3.5 db Anenna Parameers: G AZ Anenna gain (peak direciviy) 2.5 db ISSN: 1109-2742 834 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn 407 409 410 411 21m 406 495 418 493 UP 496 497 UP 425 401 DN DN 402 405 404 403 1m 1m Fig.6 Grid poin resoluion of he ervice area 33m Fig.7 Service area (he 1 s firs floor) Fig.8 Service area (he 2 nd second floor) ISSN: 1109-2742 835 Issue 8, Volume 7, Augus 2008

Chuima Prommak and Naruemon Waanapongsakorn Fig.9 Service area (he 3 rd hird floor) Fig.10 Service area (he 4 h forh floor) ISSN: 1109-2742 836 Issue 8, Volume 7, Augus 2008