Distributed predictive control of communicating and constrained systems

This paper is concerned with distributed model predictive control (DMPC) for distributed linear time invariant systems with decoupled dynamics but coupling via a common cost function and common convex state constraints. In the proposed scheme, local model predictive controllers exchange state measurements and predicted input sequences with bounded uncertainty via possibly delayed communication. A robust model predictive control (RMPC) problem is solved locally under consideration of uncertain information received from other subsystems. The local RMPCs separately optimize over nominal inputs as well as feedback policies for the delayed information and can be implemented as a tractable quadratic program. Under appropriate assumptions, the proposed scheme satisfies the coupling constraints and renders the system practically stable. Simulation results for the control of a platoon of autonomous vehicles relative to a lead vehicle illustrate the effectiveness of the approach.