Wireless Sensor Network Target Tracking Based On Particle Filter

As one of the most important technologies of the 21^st century, the wireless sensor network (WSN) has been used in many fields. One of its uses is the target tracking. Generally, the models of the target tracking are non-linear and non-Gaussian, but Kalman filters and traditional methods of data fusion can not solve non-linear and non-Gaussian problems effectively. Besides, different wireless sensor networks have different performances, such as reliability, precision and real-time performance. If we use uniform sensors in the whole network, it's quite easy to produce bottle-neck problems and the network will become noneffective.A particle set, which is randomly sampled from probability function and has corresponding weights, is introduced to approach the posterior distribution. Therefore it can handle nonlinear and non-Gaussian problems without any limits. Though the particle filter has so many advantages, there are still some existing problems, such as sample impoverishment, low performance and bad real-time performance. Since the above disadvantage of particle filter, some improved algorithms are presented to study the problems of WSN target tracking.Firstly, this project has studied WSN target tracking systems, and then introduces particle filter theory based on the analysis of non-linear filter. Since the disadvantage of multiple targets tracking algorithms, some improved algorithms, based on particle filter, U-particle filter and improved U-particle filter have been presented in this dissertation. The simulation result show that these methods perform well in WSN target tracking.