Investigation On Energy-Efficient Target Detectiion, Localization And Tracking With Wireless Sensor Networks

A wireless sensor network (WSN) is a multi-hop self-organized network system that consists of a large amount of low-cost and small-size sensor nodes with limited sensing, computing and communication capabilities. As a novel information acquisition technology, WSN has found a wide spectrum of both military and civilian applicalitons, among which lots of applications can be formulated into the target detection, localization or tracking problems. As the sensor nodes are powered by battery that is usually not rechargeable or replaceable, efficiently utilizing the node energy to improve the network liftime becomes the first consideration in the design of WSN.This dissertation studies the energy-efficient target detection, localization and tracking techniques with WSN. Firstly, the energy-efficient target detection techniques by setting optimal detection parameters are studied; Then, the energy-efficient target localization techniques by reducing the computation load, communication load and participant nodes are considered; Lastly, the energy-efficient target tracking techniques with dynamic cluster and tracking chain are investigated. The main works in the dissertation are summarized as follows:1. Study on the energy-efficient target detection techniquesTo reduce the energy consumption in target detection, the WSN can do node scheduling or detection parameter adjustment. This dissertation studies the approaches to set optimal detection parameters: detection interval and detection distance.For the setting of optimal detection interval, an improved method to solve the minimum path exposure is firstly proposed, and then the setting of detection interval is formulated as a constrained optimalization problem-minimization of the energy consumption with satisfactory minimum path exposure. A method to solve the optimalization problem is also presented. The influences of various factors on the setting of detection interval are theoretically analyzed and extensively simulated.For the setting of optimal detection distance, an improved method to solve the area coverage level is firstly proposed, and then the setting of detection distance is formulated to be a constrained optimalization problem-minimization of the energy consumption with satisfactory area coverage level. A method to solve the optimalization problem is also presented. Simulated results show that the proposed scheme consumes less energy than traditional methods.2. Study on the energy-efficient target localization techniquesTo localize a target, the WSN will consume energy in the target signal sensing, target information communication and target location estimation. This dissertation studies the approches to reduce the computation load, the communication load and the participant nodes of traditional localization methods.For the traditional energy-based target localization, this dissertation proposes two weighted least square solutions by assigning a weight to each hypersphere or hyperplane according to the noise. Only considering the large weighting hyperspheres or hyperplanes for the target location estimation, the proposed solutions can efficiently reduce the computation load with satisfactory localization accuracy. Simulated results confirm the advantages of the proposed methods.For the traditional bearing-only target localization, this dissertation proposes a node selection strategy for the case that node coordinates are imprecise. The covariance of target location estimation is firstly reduced with node coordinate error taken into account, and is then used as a metric to select the node set for target localization. Simulated results show that the proposed scheme can reduce participant sensor nodes for target localization with satisfactory localization accuracy. Furthermore, this dissertation proposes an approach to calibrate the node coordinate error with the target localization.Traditional centroid based localization is an average computation problem. This dissertation proposes a novel scheme to implement distributed average computation in WSN. In the scheme, the network of nodes operate as pulse-coupled oscillators; the data at each node is encoded into the firing time of the pulse; neighboring nodes use pulse coupling for data exchange and local computation. As the node data is reposed in a 1-bit pulse signal, the proposed scheme has much small communication load. The influences of various factors on the proposed scheme are extensively simulated.3. Study on the energy-efficient target tracking techniquesIt is expected in target tracking with WSN to arrange the sensor nodes into a tracking structure that is competent for target tracking and meanwhile consumes less energy. Motivated by this, the dissertation studies the approaches to track target with dynamic clusters and query target with tracking chain. A dynamic cluster is a tree structure consisting of the sensor nodes around the target. This dissertation emphatically studies the two key issues in target tracking with dynamic cluster, i.e. cluster construction and target estimation. For the former issue, a voting approach is developed for the cluster head selection. For the latter one, a new CPA (closest point of approach) event is defined and a target estimation algorithm is presented. Then, a dynamic cluster-based target-tracking process is presented. Simulated results show that the presented scheme has low failure probability, long network life and high tracking accuracy.A tracking chain is a bi-directional chain consisting of the sensor nodes on the target route and reserving the target spatio-temporal information. The dissertation defines the tracking chain for the first time, and proposes the approaches to construct the tracking train and to query the target with the tracking chain. The target query cost with the tracking chain is also analytically formulated. Simulated results show that the proposed scheme is energy-efficient for the cases that the target query frequency is low and the target mobility frequency is high.