Multisensor Scheduling for Remote State Estimation Over a Temporally Correlated Channel

This article studies multisensor scheduling for remote state estimation in cyber-physical systems. We consider that each sensor monitors a dynamic process and sends its data to the remote end. This article focuses on minimizing remote estimation errors over a temporally correlated communication channel. The problem is formulated as the Markov decision process (MDP) with finite-horizon cost criterion. The optimal structured policies are derived for both Markov packet dropout and finite-state Markov channel models, which can reduce computation overhead. For the infinite-horizon case, we design algorithms to address the issues of unknown channel statistics and the curse of dimensionality in the MDP, respectively. Particularly, a heuristic algorithm with linear complexity is proposed to schedule multisensor in a decentralized manner. Simulation examples are provided to verify the theoretical results.