검색 상세

Mobile Object Localization and Tracking using Wireless Infrastructure in IoT Environment

초록/요약

Internet-of-Things (IoT) is the convergence of Internet with RFID, sensors and smart objects. IoT integrates many technologies, which can enable diverse ways to obtain the positioning information of various objects. Object tracking is one of the promising IoT application which seeks to identify the position of an object over time. This dissertation focuses on mobile object localization and tracking strategies using different wireless technologies. We used WLAN based fingerprint positioning technique that uses existing Wi-Fi infrastructure to locate the mobile objects. In Wi-Fi fingerprint technique, Received Signal Strength (RSS) from different Access Points (APs) at known coordinates are stored in a database in offline phase. The stored RSS values are then matched with the RSS values recorded in real-time. The best matched value is considered the possible object location. However, the RSS values are hugely affected by geographical and other environmental factors; gives a nonlinear transformation between offline and online RSS measurements. To mitigate the undesired effects of nonlinear Wi-Fi RSSI data, we used Kernel based localized fisher discriminant analysis (KLFDA). KLFDA effectively learns distinct features in RSS data while coping with nonlinearity in data at the same time. For effective tracking of static or mobile object, accurate location estimation is the key. This dissertation’s focus is on tracking of continuous mobile objects like chemical diffusions, wildfire, and oil spills, which is a special type of mobile object tracking. Continuous objects have unique characteristics like uneven expansion, contraction, splitting, and merging and usually spreads over a very large area. Therefore, substantial number of sensing nodes monitor and observes the object. To save sparse resources of sensor networks, usually, only nodes on the boundary of the object, participate in tracking. Previous works focused on optimizing the number and size of boundary sensor nodes and boundary data and did not consider the effect of boundary node failures; which occurs likely due to uneven environmental conditions. Failure of these nodes affects the accuracy of object boundary estimation. We proposed a local failure detection and recovery mechanism which uses Voronoi based clustering technique to localize the failure detection and recovery to a Voronoi cell. Also, in previous works, the object boundary data were sent to base station even when the change in object shape is negligibly small. This process wastes the sparse resources on sending unimportant information. We proposed to avoid sending of negligible information to the base station by restricting the change in boundary shape to a certain threshold. We introduced the concept of strong and normal boundary nodes to identify the amount of change in boundary shape. This dissertation shows three main contributions; 1) accurately localizing the mobile objects in a restricted environment by mitigating the nonlinear RSS data effects through a kernel method. 2) providing robust continuous object tracking by applying Voronoi based local failure detection and recovery mechanism, and 3) restricting the selection of object boundary nodes selection criteria to avoid forwarding of useless boundary information. Proposed schemes provide robust, efficient, and accurate continuous boundary localization and tracking to support real time location and tracking applications.

more

목차

CHAPTER 1. INTRODUCTION 1
1.1 Motivation 1
1.2 Contributions of this Dissertation 2
1.3 Organization of this Dissertation 4
CHAPTER 2. BACKGROUND 6
2.1 Internet-of-Things 6
2.2 Object Tracking 7
2.2.1 Object Tracking Techniques 9
2.3 Object Localization 12
2.3.1 WLAN Based Object Localization 13
CHAPTER 3. CONTINUOUS MOBILE OBJECT TRACKING IN FAILURE-PRONE NETWORKS 19
3.1 Overview 19
3.2 Related Work 23
3.2.1 Single and Multiple Object Tracking 24
3.2.2 Continuous Object Tracking 25
3.2.3 Node Failure Detection and Recovery 28
3.3 Continuous Object Tracking for Failure-Prone Wireless Sensor Networks 29
3.3.1 Network Model 29
3.3.2 Post-Deployment Network Clustering Using a Voronoi Diagram 30
3.3.3 Continuous Object Boundary Detection and Tracking 33
3.4 Node Failure Detection and Recover 40
3.4.1 Failure Detection 40
3.4.2 Failure Recovery 42
3.5 Performance Evaluation 43
3.5.1 Simulation Environment 43
3.5.2 Simulation Results 44
3.6 Chapter Summary 55
CHAPTER 4. KERNEL BASED FEATURE-EXTRACTION METHOD FOR POSITIONING ACCURACY IMPROVEMENT 56
4.1 Overview 56
4.2 Related Work 59
4.3 Kernel Based Fisher Discriminant Analysis Scheme 63
4.3.1 Overview of the Proposed Scheme 63
4.3.2 Feature Generation 65
4.3.3 Applying Kernel Method on Local Fisher Discriminant Analysis 67
4.3.4 Choice of Kernel 68
4.3.5 Outlier Detection 69
4.4 Performance Evaluation 71
4.4.1 Experimental Data and Setup Description 71
4.4.2 Analysis of Results 72
4.5 Chapter Summary 79
CHAPTER 5. CONCLUSION 80

more
반출 Meta View 목록