Distributed Consensus-Based Multitarget Filtering and Its Application in Formation-Containment Control

This paper studies a distributed multitarget filtering problem for a sensor network where each sensor can obtain the measurements and system information of some targets while having no knowledge of others. To estimate the states of all targets, a consensus Kalman information filtering algorithm with an adaptive and finite-time matrix-weighted consensus strategy is proposed. When the communication network is strongly connected and the sensing network is time-varying while being always collectively observable, it is proved that if the targets' system matrices are time invariant, the mean-square estimation errors of the sensors are bounded for any number of consensus iterations. If the targets' system matrices are time varying and the number of the consensus steps per information filtering is larger than the diameter of the communication topology, the mean-square estimation errors of the sensors are also bounded. When each sensor is intermittently activated to observe the targets and the network does not remain collectively observable, an allowable lower bound of detection probability is derived to guarantee the stochastic boundedness of the estimation errors. Then, the filtering algorithm is applied to design a distributed containment controller for multiple agents to encircle multiple planar heterogeneous dynamic targets. Finally, simulation examples are given to illustrate the effectiveness of the algorithms.