Neuro-fuzzy admission control in mobile communications systems

Transcription

1 University of Wollongong Research Online University of Wollongong Thesis Collection University of Wollongong Thesis Collections 2005 Neuro-fuzzy admission control in mobile communications systems Raad Raad University of Wollongong, Recommended Citation Raad, Raad, Neuro-fuzzy admission control in mobile communications systems, PhD thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, Research Online is the open access institutional repository for the University of Wollongong. For further information contact the UOW Library:

2 NOTE This online version of the thesis may have different page formatting and pagination from the paper copy held in the University of Wollongong Library. UNIVERSITY OF WOLLONGONG COPYRIGHT WARNING You may print or download ONE copy of this document for the purpose of your own research or study. The University does not authorise you to copy, communicate or otherwise make available electronically to any other person any copyright material contained on this site. You are reminded of the following: Copyright owners are entitled to take legal action against persons who infringe their copyright. A reproduction of material that is protected by copyright may be a copyright infringement. A court may impose penalties and award damages in relation to offences and infringements relating to copyright material. Higher penalties may apply, and higher damages may be awarded, for offences and infringements involving the conversion of material into digital or electronic form.

3 Neuro-Fuzzy Admission Control in Mobile Communications Systems A thesis submitted in fulfillment of the requirements for the Award of the degree of Doctor of Philosophy From University of Wollongong by Raad Raad, BE (Hon 1). School of Electrical, Computer and Telecommunications Engineering. February 2005.

4 To my family 2

5 Thesis Certification I, Raad Raad, declare that this thesis, submitted in partial fulfillment of the requirements for the award of Doctor of Philosophy, in the School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, is wholly my own work unless otherwise referenced or acknowledged. This document has not been submitted for qualifications at any other academic institutions. Signed Raad Raad February 15,

6 Table of Contents List of Tables 7 List of Figures.. 8 List of Abbreviations 12 Abstract 14 Acknowledgments. 16 Prologue Introduction Background Thesis Outline Contributions Publications Overview of Cellular Mobile Networks and Technologies Outline Cellular Mobile Wireless Networks Fundamental Concepts and Issues The Cellular Concept The Handover Process Capacity and Grade of Service in Mobile Wireless Networks A Review of Existing and Envisioned Mobile Wireless Networks AMPS GSM Third and Fourth Generation Cellular Systems Cellular WLAN IEEE and Wireless ATM Ad-Hoc Networks Connection Admission Control A Review Multi-Service Connection Admission Control Admission Control in the Literature Conclusion Fuzzy and Neuro-Fuzzy Logic Systems and Application to Cellular Mobile Communications Outline Introduction Fuzzy Logic: Rules and Concepts The Structure of a General Fuzzy Logic System The Fuzzy Logic Inference Algorithm Neuro-Fuzzy Systems Neuro-Fuzzy Controllers with hybrid structure parameter learning Fuzzy Logic Applications to Networking and Cellular Systems Fuzzy Logic and Cellular Networks Fuzzy Logic and Handover Control Neuro-Fuzzy Logic and Admission Control

7 Table of Contents Application of Fuzzy logic and Neural Networks to a Number of Telecommunications Problems Conclusion Fixed Bandwidth Reservation in Multi-Service Mobile Networks Outline Introduction Multi-service Mobile Network Model and Analysis Simulation Model Analytical and Simulation Results The Analytical Approximation Micro and Pico Cellular Case Studies and Analytical Approximations Analytical Approximations and Improvements Sensitivity Study Traffic Ratio Study Cell Dwell Time Sensitivity Conclusion Optimal Utilisation in Multi-Service Mobile Networks Outline Preliminary Study and Discussion Effective Channel Utilization Case Studies with Different Traffic Ratios Utilization when class 2 = 20 kb/sec Utilization when class 2 = 50 kb/sec Utilization when class 2 = 80 kb/sec Utilization with Higher Ratios Utilization Versus the Class Ratio Optimum Utilization and Three Traffic Classes Utilization Performance and Cell Capacity Three Traffic Classes and Capacity Two Class Case Study and Capacity Optimality and multi-service mobile A Study of the Limits of the Mobile Multi-Service Utilization Conclusion Fuzzy Logic Admission Control in Micro-Cellular Mobile Networks Outline Adaptive Fuzzy Bandwidth Reservation Velocity Based Fuzzy Logic Admission Control Velocity Based Admission Control The Fuzzy Rule Set The Admission Control Surface Simulation Results for Velocity Based Admission Control Dynamic Performance of FCAC and Fixed Reservation Utilization Study of FCAC and Fixed Reservation Utilization Study of FCAC and FR with Variable Load Conclusion Neuro-Fuzzy Dynamic Bandwidth Reservation in Micro-Cellular Mobile Networks Outline Introduction Adaptive Channel Reservation Using A Neuro-Fuzzy Controller Neuro-Fuzzy Controller Structure Micro-Cellular Network Model Using NFC with Non-Exponential Cell Dwell and Call Holding Times Simulation Results Testing the NFC against the Exponential Model Applying the NFC to Lognormal Cell Dwell Times Applying the NFC to Gamma and Lognormal Cell and Call Holding Times 201

8 Table of Contents Automated Fine Tuning of the NFC Conclusion Conclusions and Recommendations for Further Research Conclusions Recommendations for Further Research Neuro-Fuzzy Logic Controller in Multi-service Mobile Networks Mobility Model Selection Application to Current Wireless Network...232

9 List Of Tables TABLE 6.1. THE FUZZY RULE SET FOR FUZZY ADMISSION CONTROL TABLE 6.2. THE CRITICAL SET OF RULES

10 List of Figures FIGURE 2.1. AN EXAMPLE MOBILE CELLULAR NETWORK WITH CHANNEL REUSE FIGURE 2.2. THE HANDOFF PROCESS [RAP2000]...33 FIGURE 2.3. TOPOLOGY OF GSM [HEI1998]...40 FIGURE 2.4 THE EFFECTIVE BANDWIDTH CONCEPT (ADOPTED FROM [BERG1998]) FIGURE 2.5. CONNECTION ADMISSION REGION FOR TWO CONNECTION TYPES...54 FIGURE 2.6. STATE-SPACE DIAGRAM OF ADMISSION REGION FOR TWO TYPES OF CONNECTIONS FIGURE 2.7. STATE SPACE DIAGRAM FOR TRUNK RESERVATION WITH A SINGLE CLASS OF TRAFFIC...57 FIGURE 2.8. STATE SPACE DIAGRAM FOR TWO CLASSES OF TRAFFIC. CLASS 2 IS RESTRICTED TO PART OF THE BANDWIDTH...58 FIGURE 3.1 CRISP AND FUZZY SETS...65 FIGURE 3.2 GENERAL STRUCTURE OF A FUZZY SYSTEM...66 FIGURE 3.3 GENERIC ADAPTIVE FUZZY SYSTEM FIGURE 3.4 FUZZY LOGIC INFERENCE ALGORITHM...68 FIGURE 3.5 EXAMPLES OF FUZZY MEMBERSHIP FUNCTIONS...69 FIGURE 3.6 FUZZY REASONING METHODS [LIN1996]...70 FIGURE 3.7 AN EXAMPLE OF THE MAMDANI S DIRECT METHOD [MAT1999]...72 FIGURE 3.8 AN EXAMPLE OF THE TAKAGI AND SUGENO S FUZZY MODELING METHOD [MAT1999]...73 FIGURE 3.9 NEURO-FUZZY CONTROLLER STRUCTURE FIGURE 4.1. CELL LAYOUT WITH RAP AROUND FOR EDGE CELLS...93 FIGURE 4.2. STATE SPACE DIAGRAM OF TRUNK RESERVATION MODEL...98 FIGURE 4.3. BLOCKING PROBABILITIES OF BOTH CLASSES OF TRAFFIC FOR NEW AND HANDOVER CALLS AS THE RESERVED BANDWIDTH IS INCREASED FIGURE 4.4. HANDOVER RATE AGAINST NEW CALL RATE FOR CLASS FIGURE 4.5. BOTH SETS OF RESULTS ARE FOR THE SAME UTILIZATION BUT WITH DIFFERENT CELL DWELL TIMES (HIGHER LINES ARE FOR CELL DWELL TIME OF 100 SECS) FIGURE 4.6. BLOCKING PROBABILITY WHEN THE CELL DWELL TIME IS MUCH LESS THAN THE CALL HOLDING TIME (CELL DWELL TIME = 12 SEC, CALL HOLDING TIME = 100 SEC) FIGURE 4.7. ANALYTICAL AND SIMULATION RESULTS FOR THE MICRO CELLULAR CASE WHERE THE CALL HOLDING TIME IS EQUAL TO THE CELL DWELL TIME (100 SECONDS) WITH THE FIRST APPROXIMATION FIGURE 4.8. ANALYTICAL AND SIMULATION RESULTS FOR THE MICRO CELLULAR CASE WITH SECOND APPROXIMATION APPLIED FIGURE 4.9. ANALYTICAL AND SIMULATION RESULTS FOR THE PICO-CELLULAR CASE WITH FIRST ANALYTICAL APPROXIMATION FIGURE ANALYTICAL AND SIMULATION RESULTS WITH THE SECOND ANALYTICAL APPROXIMATION APPLIED FIGURE HANDOVER CALL BLOCKING PROBABILITY FOR CLASS 2 AS THE RATIO OF CLASS 1 TO CLASS 2 LOAD IS VARIED WITH THE TOTAL LOAD KEPT AT THE OPTIMAL VALUE FIGURE NEW CALL BLOCKING PROBABILITIES FOR CLASS 2 AS THE RATIO OF CLASS 1 TO CLASS 2 IS VARIED FIGURE BLOCKING PROBABILITIES AGAINST RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 1000 SECONDS FOR CLASS 1 HANDOVER CALLS FIGURE BLOCKING PROBABILITIES AGAINST RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 1000 SECONDS FOR CLASS 2 HANDOVER CALLS FIGURE BLOCKING PROBABILITIES AGAINST RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 1000 SECONDS FOR CLASS 1 NEW CALLS FIGURE BLOCKING PROBABILITIES AGAINST RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 1000 SECONDS FOR CLASS NEW CALLSFIGURE BLOCKING 8

11 List of Figures 9 PROBABILITIES AGAINST THE RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 100 SECONDS FOR CLASS 1 HANDOVER CALLS FIGURE BLOCKING PROBABILITIES AGAINST THE RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 100 SECONDS FOR CLASS 2 HANDOVER CALLS FIGURE BLOCKING PROBABILITIES AGAINST THE RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 100 SECONDS FOR CLASS 1 NEW CALLS FIGURE BLOCKING PROBABILITIES AGAINST THE RESERVED BANDWIDTH FOR A MEAN CELL DWELL TIME OF 100 SECONDS FOR CLASS 2 NEW CALLS FIGURE 5.1. EFFECTIVE CAPACITY UTILISATION AGAINST CELL DWELL TIME FOR 10% AND 20% BANDWIDTH RESERVATION FIGURE 5.2. EFFECT OF CELL SIZE ON NEW CALL BLOCKING FOR CLASS 1 (A) AND CLASS 2 (B). 123 FIGURE 5.3. EFFECTIVE UTILIZATION WHEN PERFORMANCE CONSTRAINTS ARE APPLIED FOR CLASS 2 NEW CALLS FIGURE 5.4. EFFECTIVE UTILIZATION AS RESERVED BANDWIDTH IS INCREASED FIGURE 5.5.EFFECTIVE CHANNEL UTILISATION AGAINST RESERVED BANDWIDTH FOR A RANGE OF CELL DWELL TIMES FIGURE 5.6. NEW AND HANDOVER CALL BLOCKING PROBABILITIES FOR CLASSES 1 AND FIGURE 5.7. BANDWIDTH RESERVED AGAINST THE BANDWIDTH REQUIREMENT OF CLASS 2 THAT RESULT IN THE MAXIMUM EFFECTIVE BANDWIDTH UTILIZATION FIGURE 5.8. EFFECTIVE UTILIZATION FOR 3 CLASS SCENARIO FIGURE 5.9. GRAPHS SHOWING THE EFFECTIVE BANDWIDTH UTILIZATION FOR A RANGE OF CELL DWELL TIMES FIGURE EFFECTIVE UTILIZATION AGAINST RESERVED BANDWIDTH AND CELL SIZE FIGURE CORRESPONDING CLASS BLOCKING PROBABILITIES FOR CLASS 1 TO CLASS 2 RATIO OF 1 TO 2 (THE Y AXIS IS A LOG 10 SCALE) FIGURE EFFECTIVE UTILIZATION FOR TRAFFIC RATIO 1 TO 5 AND A VIEW WHEN THE CELL DWELL TIME IS 1000 SECONDS FIGURE GRAPHS OF THE CORRESPONDING BLOCKING PROBABILITIES FOR TRAFFIC RATIO OF 1 TO 5 (THE Y AXIS IS A LOG 10 SCALE) FIGURE GRAPHS OF UTILIZATION FOR TRAFFIC RATIO OF 1 TO FIGURE NEW AND HANDOVER BLOCKING PROBABILITIES FOR TRAFFIC RATIO OF 1 TO 8 (THE Y AXIS IS A LOG 10 SCALE) FIGURE GRAPH SHOWING UTILIZATION FOR THE CASE WHERE THE TRAFFIC RATIO OF 1 TO FIGURE UTILIZATION FOR THE CASE WHERE THE TRAFFIC RATIO OF 1 TO FIGURE UTILIZATION FOR TRAFFIC RATIO OF 1 TO FIGURE BLOCKING PROBABILITIES FOR RATIO OF 1 TO 20 (THE Y AXIS IS A LOG 10 SCALE) FIGURE MAXIMUM UTILIZATION VERSUS TRAFFIC RATIO FIGURE EFFECTIVE UTILIZATION FOR THE CASE WHERE THERE ARE 3 CLASSES OF TRAFFIC (10, 40 AND 80 KB/SEC) EACH APPLYING EQUAL LOAD FIGURE BLOCKING PROBABILITIES FOR THE CASE OF 3 TRAFFIC CLASSES OF 80, 40 AND 10 KB/SEC. (A), (B) AND (C) ARE HANDOVER CALLS FOR CLASSES 3, 2 AND 1 RESPECTIVELY. (D), (E) AND (F) ARE FOR NEW CALLS RESPECTIVELY (THE Y AXIS IS A LOG 10 SCALE) FIGURE THE EFFECTIVE UTILIZATION FOR THE CASE WHERE THE TRAFFIC RATIO IS 1 TO 8 AND THE TOTAL CELL BANDWIDTH IS 4000 KB/SEC FIGURE EFFECTIVE UTILIZATION FOR THE CASE OF 3 CLASSES AND TOTAL SYSTEM BANDWIDTH OF 2MB/SEC FIGURE UTILIZATION FOR THE CASE OF CLASS 1=10KB/SEC AND CLASS 2=120KB/SEC AND C=4MB/SEC FIGURE UTILIZATION FOR THE CASE WHERE CELL DWELL TIME IS SET TO 1000 SECONDS AND THE CLASS RATIO IS 1 TO 12 WITH C=6MB/SEC (A) AND C=10MB/SEC (B) FIGURE EFFECTIVE UTILIZATION FOR A 5 CLASS MOBILE MULTI-SERVICE SYSTEM WHEN THE CELL DWELL TIME IS 1000 SECONDS AND THE CALL HOLDING TIME IS 100 SECONDS FIGURE EFFECTIVE UTILIZATION FOR A 4 CLASS MOBILE MULTI-SERVICE SYSTEM WHEN ONLY THE NEW CALL BLOCKING PROBABILITY CRITERION IS MET FIGURE 6.1. FORMAL DESIGN OF FUZZY LOGIC ADMISSION CONTROLLER

12 List of Figures 10 FIGURE 6.3. MEMBERSHIP FUNCTIONS OF THE INPUT HANDOVER RATE FIGURE 6.4. MEMBERSHIP FUNCTIONS OF THE RESERVED CAPACITY FIGURE 6.5. THE RESERVED CAPACITY VERSUS THE HANDOVER RATE FIGURE 6.7. THE OUTPUT MEMBERSHIP FUNCTIONS OF THE MODIFIED FUZZY LOGIC CONTOLLER DESIGN FIGURE 6.8. THE OUTPUT OF FUZZY LOGIC RESERVATION WITH 1 INPUT OBTAINED FROM THE MODIFIED FUZZY LOGIC CONTROLLER DESIGN FIGURE 6.9. NEW AND HANDOVER BLOCKING PROBABILITIES OBTAINED FROM A NETWORK SIMULATION USING THE MODIFIED FUZZY LOGIC CONTROLLER (AFR) FIGURE COST COMPARISONS OF ADAPTIVE RESERVATION AND FIXED BANDWIDTH RESERVATION TABLE 6.1. THE FUZZY RULE SET FOR FUZZY ADMISSION CONTROL FIGURE MEMBERSHIP FUNCTIONS OF SPEED (KM/H) FIGURE MEMBERSHIP FUNCTIONS OF THE OUTPUT DECISION TABLE 6.2. THE CRITICAL SET OF RULES FIGURE ADMISSION SURFACE OBTAINED WHEN BANDWIDTH IN THE NEXT CELL IS AVAILABLE FIGURE ADMISSION SURFACE OBTAINED WHEN BANDWIDTH IN THE NEXT CELL IS NEARLY CONGESTED FIGURE ADMISSION SURFACE OBTAINED WHEN BANDWIDTH IN THE NEXT CELL IS CONGESTED FIGURE ADMISSION SURFACE OBTAINED WHEN BANDWIDTH IN THE NEXT CELL IS NOT AVAILABLE FIGURE BLOCKING PROBABILITIES FOR FUZZY CONNECTION ADMISSION CONTROL (FCAC) AND FIXED RESERVATION WITH 1 CHANNEL RESERVED (FR1) FIGURE BLOCKING PROBABILITY FOR DIFFERENT CELL SIZES FIGURE DYNAMIC BEHAVIOR OF FR1(A) AND FCAC (B) FIGURE OPTIMIZED LOADS FOR THE CONSTRAINTS OF HANDOVER BLOCKING OF AND NEW CALL BLOCKING OF FIGURE NEW CALL BLOCKING PROBABILITIES FOR VARIOUS FCAC AND FR CASES FIGURE MAXIMUM NEW CALL LOAD WHILE MAINTAINING CONSTRAINTS OF NEW CALL BLOCKING OF 0.01 AND HANDOVER CALL BLOCKING OF WITH A CONGESTION VARIATION OF 20% FIGURE THE NEW CALL BLOCKING PROBABILITY (CORRESPONDING TO THE RESULTS IN FIGURE 6.24) FIGURE HANDOVER CALL BLOCKING PROBABILITY (CORRESPONDING TO THE RESULTS IN FIGURE 6.24) FIGURE 7.1 NEURO-FUZZY CONTROLLER STRUCTURE FIGURE 7.2. A UNIFORM MOBILE NETWORK FIGURE 7.3 THE DESIGN PROCESS OF THE NFC FIGURE 7.4. FUZZY MEMBERSHIP FUNCTIONS: NEW CALL RATE BEFORE (A) AFTER (B) TRAINING FIGURE 7.5. FUZZY MEMBERSHIP FUNCTIONS FOR HANDOVER CALL RATE BEFORE (A) AND AFTER (B) TRAINING FIGURE 7.6. THE NEURO-FUZZY OUTPUT SURFACE OVER ALL THE RANGE OF NEW AND HANDOVER CALL RATES FIGURE 7.8. RELATIONSHIP BETWEEN α AND C FIGURE 7.9 OUTPUT ADJUSTMENT PROCESS FIGURE COST FUNCTION OF VARIOUS BANDWIDTH RESERVATION SCHEMES FIGURE NEW AND HANDOVER BLOCKING PROBABILITIES FOR NFC FIGURE OUTPUT SURFACES BEFORE (A) AND AFTER (B) THE FINAL ADJUSTMENT FIGURE NEW AND HANDOVER BLOCKING PROBABILITIES FOR NFC IN A TEST CELL FIGURE NEW CALL RATE MEMBERSHIP FUNCTIONS FIGURE HANDOVER CALL RATE MEMBERSHIP FUNCTIONS FIGURE OUTPUT SURFACE FOR FUZZY LOGIC ADMISSION CONTROLLER FIGURE NEW AND HANDOVER BLOCKING PROBABILITY WITH ERROR BARS FIGURE NEW AND HANDOVER BLOCKING PROBABILITY WITHOUT ERROR BARS FIGURE NEW CALL RATE MEMBERSHIP FUNCTIONS AFTER FIRST ADJUSTMENT

13 List of Figures 11 FIGURE HANDOVER CALL RATE MEMBERSHIP FUNCTIONS AFTER FIRST ADJUSTMENT FIGURE DIFFERENCE BETWEEN HANDOVER CALL RATE MEMBERSHIP FUNCTIONS BETWEEN ORIGINAL AND FIRST ADJUSTMENT FIGURE OUTPUT SURFACE FOR FUZZY LOGIC ADMISSION CONTROLLER AFTER FIRST ADJSUTMENT FIGURE DIFFERENCE BETWEEN OUTPUT SURFACES FOR FUZZY LOGIC ADMISSION CONTROLLER AFTER FIRST ADJUSTMENT FIGURE OUTPUT SURFACE FOR FUZZY LOGIC ADMISSION CONTROLLER AFTER SECOND ADJUSTMENT FIGURE DIFFERENCE BETWEEN OUTPUT SURFACES FOR FUZZY LOGIC ADMISSION CONTROLLER AFTER SECOND ADJUSTMENT FIGURE DIFFERENCE BETWEEN OUTPUT SURFACES FOR FUZZY LOGIC ADMISSION CONTROLLER BETWEEN ORIGINAL AND 2 ND ADJUSTMENT FIGURE NEW CALL RATE MEMBERSHIP FUNCTIONS AFTER SECOND ADJUSTMENT FIGURE HANDOVER CALL RATE MEMBERSHIP FUNCTIONS AFTER SECOND ADJUSTMENT FIGURE NEW AND HANDOVER BLOCKING PROBABILITY WITHOUT ERROR BARS FOR THE ORIGINAL ADMISSION CONTROLLER AND THE TWO ADJUSTMENT ALGORITHMS FIGURE DIFFERENCE BETWEEN OUTPUT SURFACES FOR FUZZY LOGIC ADMISSION CONTROLLER BETWEEN ORIGINAL AND 2 ND ALGORITHM FIGURE DIFFERENCE BETWEEN OUTPUT SURFACES FOR FUZZY LOGIC ADMISSION CONTROLLER BETWEEN ORIGINAL AND 2 ND ALGORITHM (ANOTHER VIEW)

14 List of Abbreviations ATM GoS QoS IP NN NF CAC GPRS GSM PCN PCS 2G Asynchronous Transfer Mode Grade of Service Quality of Service Internet Protocol Neural Network Neuro-Fuzzy Connection Admission Control General Packet Radio Service Global System of Mobile communications Personal Communications Networks Personal Communications Services Second Generation mobile technology (such as GSM) 2.5G Systems such as GPRS built as enhancements to 2G 3G 3GPP 3GPP2 ABR AMPS BER B-ISDN BSC BSS BTS CDMA CRC D-AMPS Third Generation mobile technology Third Generation Partnership Project (WCDMA) 2 nd Third Generation Partnership Project (CDMA2000) Available Bit Rate Advanced Mobile Phone System Bit Error Rate Broadband ISDN Base Station Controller Base Station System or Subsystem Base Transceiver Station Code Division Multiple Access Cycle Redundancy Check Digital AMPS, IS

15 List of Abbreviations 13 DECT EDGE GMSC HiPerLAN HSCSD ISDN WLAN MAC UWB VBR WCDMA Digitally Enhanced Cordless Communications Enhanced Data rates for GSM Evolution Gateway Mobile Services Switching Center High Performance radio Local Area Network High Speed Circuit Switched Data Integrated Services Digital Network Wireless Local Area Network Medium Access Protocol Ultra Wide Band Variable Bit Rate Wide band CDMA

16 Abstract This thesis examines the problem of admission control in multi-service mobile communications networks. The work is divided into two parts, chapters 2, 4 and 5 look at multi-service admission control in the mobile domain, while Chapters 3, 6 and 7 propose Fuzzy and Neuro-Fuzzy Admission Control schemes for Multi-service mobile networks. The thesis begins by examining guard channel techniques for guaranteeing better performance for handover connections with different classes of traffic. An approximate analysis for fixed multi-service networks is extended into the mobile domain. This analysis is then verified through simulation. The analysis of the mobile network is further extended to include the micro-cellular case. The simulation results and analysis lead to the development of a heuristic for the amount of bandwidth that needs to be reserved for different multi-service scenarios. A fundamental contribution of this thesis is the development of admission control schemes that apply to networks with a general cell dwell time distribution (such as lognormal and gamma distributions) that are different to the classic guard channel problem. In this thesis Fuzzy and Neuro-Fuzzy admission control approaches are examined to meet performance constraints of call blocking probabilities and to optimize wireless channel utilization. A simple Fuzzy Logic controller is first implemented that allows the admission controller to make decisions about new 14

17 Abstract 15 calls attempting to access the network. This work is then extended to a completely adaptive Neuro-Fuzzy Controller. The Neuro-Fuzzy admission controller allows for the training and adaptive learning of the arrival and service profile of calls in each cell. An analytical model is developed for use as the initial training set of the Neuro-Fuzzy admission controller. An algorithm that allows for the updating of the neural network as data is collected is also developed. The proposed model allows the Neuro-Fuzzy admission controller to handle different cell dwell time distributions. This makes the admission controller more applicable to more realistic cell distributions and highly adaptive to coverage areas which are not covered by current models. Finally, the same Neuro-Fuzzy logic controller is extended to allow more variables to be considered (such as user velocity and cell dwell time). The results obtained show that the original Neuro-Fuzzy controller handles the simulated scenarios without the need for more complex extensions.

18 Acknowledgments This thesis has been a long journey for me completed over many years and there many people that I wish to acknowledge in this short space. First and foremost I would like to acknowledge the contributions of Professor Joe Chicharo who has encouraged me and lead me at every turn in the right direction. His tireless efforts resulted in this work being finally completed. I would also like to thank Professor Eryk Dutkiewicz for his contributions over the years. A special thank you goes to my wife for being so patient over the last few years. Without her support and patience this work would have been more difficult. A final thank you to the rest of my family who have been very supportive over the years. 16

19 Abstract 17 Prologue This work was supported by Motorola Australia Research Laboratories where I have worked for a number of years. During this period along with this thesis, I managed to work on a number of different and interesting research projects mainly in the emerging area of sensor networks and Quality of Service in IEEE e. In those 3 years, I filed seven patents in relation to power saving, MAC protocols and quality of service as well as producing more than 20 reports and technical documents. Currently I am a research fellow at the University of Wollongong where I am continuing research into MAC protocols in sensor networks.