Distributed Kalman filter for multi-agent tracking with Markov intermittent observation

Peng Shi, Guoyan Huang

Abstract: We consider the problem of multiple robots cooperatively tracking several moving targets using onboard sensors. Existing methods may diverge under Markov intermittent observations. We first model the problem as a linear time-varying system with a Markov process, and develop an optimal Kalman consensus filter by minimizing the estimation error. To obtain a scalable algorithm, we propose an approximate algorithm and present a sufficient condition for stability. The effectiveness and robustness with respect to network topology and Markov chain are verified through several experiments.