Synthesis of Networked Switching Linear Decentralized Controllers

This paper proposes an approach based on linear matrix inequalities for synthesizing a set of decentralized regulators for discrete-time linear systems subject to input and state constraints. Measurements and command signals are exchanged over a sensor/actuator network, in which some links are subject to packet dropout. The resulting closed-loop system is guaranteed to asymptotically reach the origin, even if every local actuator can exploit only a (possibly time-varying) subset of state measurements. A model of packet dropout based on a finite-state Markov chain is also considered to exploit available knowledge about the stochastic nature of the network. For such model, a set of decentralized switching linear controllers is synthesized that guarantees mean-square stability of the overall controlled process under packet dropout and soft input and state constraints.