Consensus for Multiagent Systems With Input Delay and Observation Packet Loss

Abstract

In this article, the mean-square consensus problem for linear discrete multiagent systems over both undirected and directed graphs is explored. The agent dynamics in the network is influenced by both input delay and observation packet loss. The packet loss probabilities are allowed to be nonidentical for every agent. An optimal predictor-based protocol is developed, and conditions ensuring the mean-square consensus are provided, which are expressed in terms of system dynamics, graph parameters, and the maximal packet loss probability. It is highlighted that consensus condition upon the maximal packet loss probability is also necessary, and the effect of input delay can be completely compensated by the proposed approach. As a special case, the mean-square consensus value for the first-order integrator systems is proved to be dependent on the initial state values. A numerical simulation is provided to validate the correctness of the derived results.