不完全量测下基于信息一致性的分布式容积卡尔曼滤波算法

Information Consensus-based Distributed Cubature Kalman Filtering Algorithm with Intermittent Observations

针对不完全量测条件下分布式火控系统中的非线性目标跟踪问题,为提高跟踪系统的估计精度并保证各探测单元估计结果的一致性,提出一种基于信息一致性的分布式容积卡尔曼滤波(ICDCKF)算法。针对非线性系统,给出不完全量测下的改进容积卡尔曼滤波。考虑到各探测单元间局部估计信息的相关性,该算法首次将协方差交叉方法应用于非线性一致性滤波算法,提高互协方差未知情形下分布式融合的估计精度。特别地,为确保算法的可行性,给出不完全量测情形下,ICDCKF算法估计结果收敛的条件,并从理论上严格证明在该条件下ICDCKF算法可以保证估计方差的有界性。ICDCKF算法应用于一类光电跟踪网络,与现有CKFI算法、CKF_CI算法、KCF_Me算法对比分析表明:ICDCKF算法在保证各探测单元估计结果一致性的同时,大幅度提高了跟踪系统的估计精度。

For nonlinear target tracking of distributed fire control systems with intermittent observations,an information consensus-based distributed cubature Kalman filter(ICDCKF)algorithm is presented to improve the estimation accuracy of tracking system and ensure the consistency of estimated results of each sensor node.The modified cubature Kalman filter with intermittent observations is presented for nonlinear systems.Considering the correlation between local estimation information of sensor nodes,the covariance intersection method is firstly utilized in the nonlinear consensus filtering algorithm,and the estimation accuracy is improved under the condition of unknown cross-covariances.Specially,the condition for guaranteeing the convergence of estimated results of ICDCKF slgorithm with intermittent observations is derived for the feasibility.In this condition,the boundedness of the estimation covariance is proved strictly in theory.The proposed ICDCKF algorithm is applied to an electo-optical sensor network.The results show that the proposed ICDCKF algorithm can greatly improve the estimation accuracy of tracking system with the consensus estimation compared with CKFI,CKF_CI and KCF_Me algorithms.