Research On Distributed Kalman Filter Method For Target Tracking

Target tracking and data fusion has been an important part of modern military control and decision-making system. Multi-sensor is widely used in target tracking for its general and full information, meeting the demand for information of target tracking system. Centralized data processing and distributed data processing are the two main methods of multi-sensor information processing. And the distributed data processing is currently the most popular method of information fusion with the advantages of simply calculation, flexible network structure and fine reliability.In the first part, several common target motion models and the corresponding mathematical equations are introduced. According to estimation theory of target tracking,several nonlinear filtering algorithms are introduced, especially the unscented Kalman filtering process.In the second part, the paper introduces distributed filtering method for multi-sensor target tracking information processing. Combining with nonlinear UKF algorithm, a distributed UKF is proposed in order to solve the nonlinear target tracking problem. By the contrast with centralized filtering algorithm, simulation experiments prove the effectiveness of the proposed algorithm. When noise of the target tracking mathematical model is unknown,the estimation results have large error, even probably leading to filtering divergence. The Sage-Husa algorithm is an effective solution of this problem for linear system. This algorithm gives the real-time noise estimation in the process of filtering. Many researches are made to combine nonlinear filtering ,method with real-time noise estimation algorithm for nonlinear field. So, a distributed adaptive UKF method with real-time noise estimation is proposed in the paper to estimate the target state with multi-sensor tracking system while the noise parameters are unknown. In the practical application, there may be situations that some sensors were broken, which would have impact on tracking precision. Towards to the problem,researches on adaptive distributed algorithm with variable network structure are made by the paper, and two ways to vary the structure are proposed. This adaptive distributed UKF method with variable network structure disconnects information of broken sensor, by which the algorithm can handle with estimation problems in this situation and improve the reliability and fault-tolerance of the system simultaneously.In the third part, the paper makes researches on multi-sensor and multi-target tracking problem, of which most important part is data associated method. The paper makes modifications on JPDA algorithm for its disadvantages, based on which a distributed UKF with improved JPDA method is proposed for the application of multi-sensor and multi-target tracking problem.