Networked Kalman filtering with combined constraints of bandwidth and random delay

Abstract

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.