Stability and Regret bounds on Distributed Truncated Predictive Control for Networked Dynamical Systems

Abstract: This work is primarily concerned about the distributed control of networked linear timeinvariant (LTI) systems. In particular, we propose a truncated predictive control algorithm based on $\kappa$-hop neighbourhoods of the agents of the networked system. We establish stability and regret bounds for the proposed algorithm, which shows that the regret decays exponentially when the temporal prediction horizon $k$ and the spatial radius $\kappa$ increases.