DISTRIBUTED TARGET TRACKING IN WIRELESS SENSOR NETWORK

Abstract

Target tracking is an important element of surveillance, guidance, or avoidance systems, whose role is to determine the position, and movements of target. ,Filtering is a way to achieve this by estimating the state of the system from noisy measurements as and when they arrive. Bayesian filtering is the most commonly used framework for tracking applications. The problem of target tracking in wireless sensor networks (WSN's) is recursive Bayesian sequential estimation problem. The central idea to this recursive Bayesian estimation is to calculate the probability density function of the state vector conditioned on the available measurements. It is known that Kalman filter provides an optimal solution to the Bayesian sequential problem for linear/Gaussian systems. However for nonlinear problems, Kalman filter cannot provide the optimal solution. For weekly non-linear systems extended Kalman filter can provide good performance but fails for highly non-linear systems. Particle filtering is an emerging and powerful methodology particularly useful in dealing with nonlinear and non-Gaussian problems based on the concept of sequential importance sampling and Bayesian theory. Distributed implementation allows the system to scale seamlessly to larger tracking regions without encountering problems of network congestion and short lifetimes. In this Dissertation, we shall consider Kalman filter and its distributed implementation. Distributed implementation of extended Kalman filter is discussed for non-linear Gaussian systems. Particle filter provides a numerical solution for non-linear and non-Gaussian systems. We discuss and compare the distributed versions of the above algorithms and their variants using various strategies for linear and non-linear models. We shall also propose a distributed algorithm using modified diffusion strategy to improve the performance of tracking in WSN. MATLAB is used for simulations.