Research On Distributed Consensus Based Target Tracking Technology In Wireless Sensor Networks

In recent years, proposing distributed filter algorithms for targeting and tracking based on consensus algorihtms in sensor networks, with high quality of performance, is becoming a popular research field. This paper focus es on combining gossip algorithms, belonging to consensus algorithms, and Kalman filter algorithms, pertaining to estimation algorithms, to research on distributed consensus based tracking in wireless sensor networks.Firstly, related research results in distributed tracking in wireless sensor networks are classified and arranged, especially that consensu s based Kalman filter algorithms recently published are elaborately analysed with their advantages and disadvantages. Following with the sufficient research on the related work, the efforts are made towards improving the important distributed Kalman filter recently come up with termed Kalman Consensus Filter(KCF), by combining gossip algorithms with Kalman filter algorithms resulting in a Gossip based Distributed Kalman Filter(GDKF). In GDKF, the informaiton across the network can be fused at each sensor by gossiping among neighboring sensors, which breaks the restriction of limited sensing and communication ranges on the performance of KCF, resulting in a distributed Kalman filter algorithm with higher tracking accuracy, higher consensus accuracy and faster convergence speed compared with KCF.Secondly, the content and rationale of KCF are studied in detail, whose application steps in sensor networks are given. The backwards of KCF are analysed, that intuitively KCF does not perform well for a minority of nodes and their neighbors that make relatively poor observations due to conditions on the sensing and communication ranges. To solve such problem, GDKF is proposed by exploiting the advantages of both gossip algorithms and KCF. The content and rationale of GDKF are given, with 2 realization algorithms provided, i.e. Pairwise Gossip based Distributed Kalman Filter(PG-DKF) and Broadcast Gossip based Distributed Kalman Filter(BG-DKF). The according application steps of PG-DKF and BG-DKF are given respectively.Finally, the performance of GDKF is theoretically analysed and has been verified by simulations compared with KCF. On the one hand, the global asymptotic stability of GDKF is proved by the Second Law of Lyapunov Stability, with the error reducing rate of GDKF deduced theoretically, which is faster than that of KCF. On the other hand, simulations in networks of heterogenous sensors with limited sensing and communication ranges are made with GDKF and KCF for tracking a manuevering target in a restricted area. The experiment results confirm the correctness of theoretical analysis, that the proposed GDKF performs with higher tracking accuracy, higher consensus accuracy and faster convergence speed, compared with KCF.