Research On Distributed Kalman Filter Collaborative Algorithm

The wireless sensor network(WSN)has been rapidly developing along with the advancement of science and technology at present.It is gradually applied to the various fields,including target tracking,precision agriculture,commercial manufacture,military operations and so on.In this paper,three kinds of distributed Kalman filtering algorithm were summarized in target tracking of wireless sensor network,including consensus Kalman filtering algorithm and two kinds of Kalman filtering algorithm based on diffusion strategies and multi-level Kalman filtering.The principle and experimental results of these algorithms were described as following.1.The algorithm of target tracking was summarized based on distributed Kalman filter,and the development histories of Kalman filtering,algorithm and basic cycle process were described.At the same time,the problems in practical application have been analyzed and the countermeasures were pointed out.2.The problems of distributed Kalman filtering have been summarized in a set of nodes estimating the state of a linear dynamic system in a collaborative manner.Our focus was on diffusion strategies,where nodes communicated with their direct neighbors only,and the information was diffused across the network through a sequence of Kalman iterations and data-aggregation.Distributed strategies were an attractive alternative,since they were in general more robust,might require fewer communications,and require a less amount of energy for communications,as well as allow parallel processing.3.Aiming at the average-rule diffusion strategy of WSN,the relative-degree rule algorithm was discussed theoretically and experimentally.The simulation results were compared with the theoretical expressions,and then the convergence of the proposed algorithms was analyzed,and mean and mean-square analysis were provided,which means that the proposed approach outperforms existing techniques.4.Considering distributed network structure of multi-level network,multi-level distributed kalman filtering algorithm is proposed in this paper.Combined with the relative-degree rule and compared with the single-level distributed algorithm,multi-level distributed network improved the robustness of the system,and had the wider coverage and higher accuracy.