Networked Kalman filtering with combined constraints of bandwidth and random delay  

作　　者：GE QuanBo WEN ChengLin XU TingLiang 

机构地区：[1]School of Automation, Hangzhou Dianzi University, Hangzhou 310018, China [2]State Key Laboratory of Industrial Control Technology, Zhejiang University, Hangzhou 310027, China

出　　处：《Science China(Information Sciences)》2012年第10期2343-2357,共15页中国科学（信息科学）（英文版）

基　　金：supported by National Nature Science Foundation of China (Grant Nos. 61172133,91016020,60934009,61002018);China Postdoctoral Science Foundation (Grant No. 20100471727);Zhejiang Nature Science Foundation (Grant No. Y1100227);Ningbo Nature Science Foundation (Grant No. 2011A610-180)

摘　　要：Kalman filter is a powerful tool in target tracking and self-localization across wireless sensor networks with many constraints. This paper considers the filters design for networked systems with combined constraints of bandwidth and random delay, and proposes a kind of universal networked Kalman estimator for given linear time invariant （LTI） or determinate parameters （DP） systems, which are also called linear and determinate parameters （LDP） systems jointly. Firstly, two modeling ways to treat bit quantization error vector are given and the corresponding LDP systems taken after bit quantization are established. Secondly, we introduce an equivalent weighted summation form of the conventional linear Kalman filter, which fully uses two essential characteristics, namely offline computation property from the LDP systems and linear weighted summation property from the linear minimum mean square error estimate. Finally, by adopting the replacement and innovation compensation operations, a kind of integrated networked estimator, involving two estimation algorithms with different modeling ways to bit quantization error actually, is presented based on the weighted summation filter for random delay systems. The proposed universal networked estimation algorithms have some outstanding advantages such as extensive application, concise algorithm structure, high estimate accuracy, and good running performance. Several examples are demonstrated to validate the proposed networked estimation algorithms.Kalman filter is a powerful tool in target tracking and self-localization across wireless sensor networks with many constraints. This paper considers the filters design for networked systems with combined constraints of bandwidth and random delay, and proposes a kind of universal networked Kalman estimator for given linear time invariant （LTI） or determinate parameters （DP） systems, which are also called linear and determinate parameters （LDP） systems jointly. Firstly, two modeling ways to treat bit quantization error vector are given and the corresponding LDP systems taken after bit quantization are established. Secondly, we introduce an equivalent weighted summation form of the conventional linear Kalman filter, which fully uses two essential characteristics, namely offline computation property from the LDP systems and linear weighted summation property from the linear minimum mean square error estimate. Finally, by adopting the replacement and innovation compensation operations, a kind of integrated networked estimator, involving two estimation algorithms with different modeling ways to bit quantization error actually, is presented based on the weighted summation filter for random delay systems. The proposed universal networked estimation algorithms have some outstanding advantages such as extensive application, concise algorithm structure, high estimate accuracy, and good running performance. Several examples are demonstrated to validate the proposed networked estimation algorithms.

关 键 词：Kalman filter LDP systems LMMSE limited bandwidth random delay 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置] TN713[自动化与计算机技术—控制科学与工程]