Research On Collaborative Target Tracking Based Gaussian Particle Filtering

Wireless Sensor Networks (WSN) is considered as one of the most significant emerging technologies in21st century. And target tracking is an important application area of WSN. Its main purpose is to estimate the target trajectories by measuring the distance or the angle between target and sensor node. Target tracking has been widely used in the military environment, health care, transportation, etc.In this thesis, characteristics of wireless sensor networks and target tracking situation are introduced. And the target tracking algorithm, specific tracking strategies and evaluation indicators of the tracking are systematically analyzed. On this basis, a multi-sensors tracking program with Gaussian particle filter is proposed. In this program, a dynamic clustering technology is used to choose sensors in real-time for tracking target. The program can reduce the network energy consumption while maintaining the tracking accuracy.Since the resource of sensor networks is limited, choosing Gaussian particle filtering which does not require resampling to estimate the state of target, greatly reduces the computation complexity of nonlinear tracking, while selecting covariance intersection(CI) algorithm to achieve a multi-sensor data fusion. Moreover, selecting cluster members is based on the position predicted by GPF, and using partile swarm optimization method to select the cluster head for data collection and forwarding. The proposed program does not wake all the sensors in network, and it can decrease the energy consumption of network transmission. At the same time, it avoids network congestion caused by large amounts of data transmission, too much control signals, radio channel congestions and so on.Finally, a series of simulation examples are given for proposed target tracking program. The simulation results show that proposed tracking program can reduce network energy consumption while maintaining a better tracking accuracy.