Target Tracking Using Wireless Sensor Networks
Abstract
Tracking of targets in remote inaccessible areas is an important application of Wireless Sensor Networks (WSNs). The use of wired networks for detecting and tracking of intruders is not feasible in hard-to-reach areas. An alternate approach is the use of WSNs to detect and track targets. Furthermore, the requirements of the tracking problem may not necessarily be known at the time of deployment. However, issues such as low onboard power, lack of established network topology, and the inability to handle node failures have limited the use of WSNs in these applications. In this dissertation, the performance of WSNs in remote surveillance type of applications will be addressed through the development of distributed tracking algorithms. The algorithm will focus on identifying a minimal set of nodes to detect and track targets, estimating target location in the presence of measurement noise and uncertainty, and improving the performance of the WSN through distributed learning. The selection of a set of sensor nodes to detect and track a target is first studied. Inactive nodes are forced into `sleeping' mode to conserve power, and activated only when required to sense the target. The relative distance and angle of the target from sensor nodes are used to determine which of the sensors are needed to track the target. The effect of noisy measurements on the estimation of the position of the target is addressed through the implementation of a Kalman filter. Contrary to centralized Kalman filter implementations reported in the literature, implementation of the distributed Kalman filter is considered in the proposed solution. Distributed learning is implemented by passing on the knowledge of the target, i.e. the filter state and covariance matrix onto the subsequent node running the filter. The problem is mathematically formulated, and the stability and tracking error of the proposed strategy are rigorously examined. Numerical examples are then used to demonstrate the utility of the proposed technique. It will be shown by mathematical proofs and numerical simulation in this dissertation that distributed detection and tracking using a limited number of nodes can result in efficient tracking in the presence of measurement noise. Furthermore, minimizing the number of active sensors will reduce communication overhead and power consumption in networks, improve tracking efficiency, and increase the useful life span of WSNs.
Collections
- OU - Dissertations [9305]