DESIGN OF INDOOR POSITIONING SYSTEMS BASED ON LOCATION FINGERPRINTING TECHNIQUE by Kamol Kaemarungsi B. Eng., King Mongkut's Institute of Technology at Ladkrabang, Thailand, 1994 M. S. in Telecommunications, University of Colorado at Boulder, 1999 Submitted to the Graduate Faculty of the School of Information Science in partial ful¯llment of the requirements for the degree of Doctor of Philosophy University of Pittsburgh 2005 UNIVERSITY OF PITTSBURGH SCHOOL OF INFORMATION SCIENCE This dissertation was presented by Kamol Kaemarungsi It was defended on February 16, 2005 and approved by Prashant Krishnamurthy, Ph. D., Assistant Professor Richard Thompson, Ph. D., Professor Joseph Kabara, Ph. D., Assistant Professor Ching-Chung Li, Ph. D., Professor Panos Chrysanthis, Ph. D., Professor Dissertation Director: Prashant Krishnamurthy, Ph. D., Assistant Professor ii Copyright °c by Kamol Kaemarungsi 2005 iii DESIGN OF INDOOR POSITIONING SYSTEMS BASED ON LOCATION FINGERPRINTING TECHNIQUE Kamol Kaemarungsi, PhD University of Pittsburgh, 2005 Positioning systems enable location-awareness for mobile computers in ubiquitous and per- vasive wireless computing. By utilizing location information, location-aware computers can render location-based services possible for mobile users. Indoor positioning systems based on location ¯ngerprints of wireless local area networks have been suggested as a viable so- lution where the global positioning system does not work well. Instead of depending on accurate estimations of angle or distance in order to derive the location with geometry, the ¯ngerprinting technique associates location-dependent characteristics such as received signal strength to a location and uses these characteristics to infer the location. The advantage of this technique is that it is simple to deploy with no specialized hardware required at the mobile station except the wireless network interface card. Any existing wireless local area network infrastructure can be reused for this kind of positioning system. While empirical results and performance studies of such positioning systems are pre- sented in the literature, analytical models that can be used as a framework for e±ciently designing the positioning systems are not available. This dissertation develops an analytical model as a design tool and recommends a design guideline for such positioning systems in order to expedite the deployment process. A system designer can use this framework to strike a balance between the accuracy, the precision, the location granularity, the number of access points, and the location spacing. A systematic study is used to analyze the location ¯ngerprint and discover its unique properties. The location ¯ngerprint based on the received signal strength is investigated. Both deterministic and probabilistic approaches of location iv ¯ngerprint representations are considered. The main objectives of this work are to predict the performance of such systems using a suitable model and perform sensitivity analyses that are useful for selecting proper system parameters such as number of access points and minimum spacing between any two di®erent locations. Keywords: pattern classi¯cation, performance, position location system, system design, wireless local area network. v TABLE OF CONTENTS PREFACE ......................................... xv I. INTRODUCTION ................................ 1 A. INTRODUCTION TO THE STUDY ..................... 1 B. BACKGROUND OF INDOOR POSITIONING SYSTEMS ......... 2 C. INDOOR POSITIONING SYSTEMS BASED ON LOCATION FINGER- PRINTING ................................... 6 D. APPROACHES AND CONTRIBUTIONS .................. 8 E. ORGANIZATION ................................ 10 II. LITERATURE REVIEW ............................ 11 A. COMMON COMPONENTS OF INDOOR POSITIONING SYSTEMS ... 11 B. TAXONOMY OF INDOOR POSITIONING SYSTEMS ........... 13 1. Sensing Technologies ............................. 13 2. Measurement Techniques .......................... 14 3. Location System Properties ......................... 17 C. RELATED INDOOR POSITIONING SYSTEMS ............... 17 D. INDOOR POSITIONING SYSTEMS USING WIRELESS LANS AND LO- CATION FINGERPRINTING ......................... 20 1. Indoor Environment ............................. 21 2. Location Fingerprint ............................. 22 3. Location Estimation Algorithm ....................... 25 a. Nearest Neighbor Methods ........................ 26 b. Neural Network Methods ........................ 29 vi c. Probabilistic Methods .......................... 31 d. Support Vector Machine Methods .................... 33 4. Summary of Existing Indoor Positioning Performance .......... 35 E. CONCLUSIONS ................................. 37 III. PROPERTIES OF RECEIVED SIGNAL STRENGTH ......... 40 A. MEASUREMENT SETUP ........................... 40 1. Experimental Design ............................. 42 B. COLLECTING MEASUREMENT OF RECEIVED SIGNAL STRENGTH 49 1. User's E®ect ................................. 49 a. User's Body ................................ 49 b. User's Orientation ............................ 50 2. Make of Wireless Card ............................ 51 a. Impact of Quantization of RSS Values by Wireless Cards ....... 56 C. STATISTICAL PROPERTIES OF RECEIVED SIGNAL STRENGTH .. 58 1. Distribution of the Received Signal Strength ................ 58 2. Standard Deviation of the Received Signal Strength ........... 63 3. Stationarity of the Received Signal Strength ................ 64 4. Time Dependency .............................. 67 a. Temporal Received Signal Strength Outage .............. 70 5. Interference and Independence ....................... 70 6. Required Number of Samples ........................ 72 D. CAUSES OF ERROR IN LOCATION DETECTION ............ 73 1. Randomness of Received Signal Strength Patterns ............ 74 2. Separation of Location Fingerprints .................... 75 3. Separation by Path Loss in Signal Propagation .............. 80 E. SUMMARY OF ANALYSIS RESULTS .................... 82 F. IMPLICATION ON MODELING OF LOCATION FINGERPRINT .... 84 G. CONCLUSIONS ................................. 85 IV. MODELING OF THE POSITIONING SYSTEM ............. 87 A. MODEL FROM PATTERN CLASSIFICATION ............... 88 vii 1. Probabilistic Approach with Lognormal Assumption ........... 90 2. Euclidean Distance with Lognormal Assumption and Identity Covariance 94 B. PROBABILITY OF RETURNING THE CORRECT LOCATION ..... 99 1. Alternate Calculation of Probability with the Euclidean Distance .... 101 2. Extension to Multi-Location Systems .................... 103 C. PERFORMANCE EVALUATION ....................... 105 1. System Model Setup ............................. 105 2. Results of the Probability of Returning the Correct Location for a Single Neighbor ................................... 108 3. Results of the Probability of Returning the Correct Location for Multiple Neighbors ................................... 111 D. ERROR DISTANCE DISTRIBUTION .................... 112 E. GOODNESS OF ANALYTICAL AND SIMULATION MODELS ...... 115 1. Missing Signals ................................ 116 2. Non Uniform Grid Spacing ......................... 117 3. Continuity of Real Location ......................... 118 F. CONCLUSIONS ................................. 119 V. SYSTEM DESIGN AND DEPLOYMENT ................. 120 A. HIGH LEVEL SYSTEM DESIGN ....................... 120 1. System Design Issues ............................ 121 B. LOW LEVEL SYSTEM DESIGN ....................... 124 1. Algorithm Selection ............................. 124 a. Performance Comparison of Di®erent Positioning Algorithms ..... 126 2. Performance Achieving Design ....................... 128 a. Design Examples Using Proposed System Model ............ 129 C. PROTOTYPE OF INDOOR POSITIONING SYSTEM ........... 132 1. Performance Evaluation of Prototype ................... 134 2. Performance Improvement Guidelines for Prototype System ....... 138 D. HIGH LEVEL DESIGN GUIDELINES .................... 139 1. Design Decision Guidelines ......................... 139 viii E. CONCLUSIONS ................................. 139 VI. CONCLUSIONS AND FUTURE WORK .................. 141 A. CONTRIBUTIONS ............................... 142 B. FUTURE RESEARCH WORK ........................ 143 APPENDIX A. FIGURES OF RECEIVED SIGNAL STRENGTH DIS- TRIBUTION .................................... 145 APPENDIX B. TABLES OF STANDARD DEVIATION ........... 150 APPENDIX C. TABLES OF VARIANCE AND COVARIANCE ...... 153 APPENDIX D. TABLES OF CORRELATION COEFFICIENT ....... 154 APPENDIX E. FIGURES OF CORRELOGRAMS ............... 161 APPENDIX F. SUMMARY STATISTICS WITH DIFFERENT NUMBER OF SAMPLES ................................... 163 APPENDIX G. FLOWCHARTS OF INDOOR POSITIONING PROTO- TYPE ........................................ 168 APPENDIX H. LOCATION FINGERPRINTS OF PROTOTYPE'S EX- PERIMENT .................................... 171 BIBLIOGRAPHY .................................... 173 ix LIST OF TABLES 1 Summary of location stack ........................... 13 2 Properties of location systems ......................... 18 3 Parameter comparison of indoor positioning systems ............. 35 4 Performance comparison of indoor positioning systems ............ 36 5 Factors that a®ect RSS ¯ngerprint ....................... 43 6 Measurable access points on 4th floor in Information Science building ... 45 7 Access points in Hillman library ........................ 48 8 Experimental design and measurement factors ................ 48 9 Sample