A Cooperative Distributed MPC Algorithm With Event-Based Communication and Parallel Optimization

This paper investigates the use of event-based communication in a distributed model predictive control (DMPC) scheme for linear subsystems interconnected by dynamics and costs. In the proposed DMPC scheme, all subsystems optimize their local input sequences in parallel, and local iterations are performed to update the global input sequence. To reduce the load on the communication network, we propose an event-based communication protocol, in which local information is only communicated if doing so results in a sufficient improvement of the overall control performance. Based on the event generator and a distributed stopping criterion, we first establish that the scheme terminates after a finite number of iterations, and we provide bounds on the suboptimality of the solution. It is shown that the suboptimality of the scheme can be made arbitrarily small by choosing an appropriate threshold. Subsequently, a bound on the convergence rate is established. Based on this bound, parameters used in the scheme are optimized for fast convergence. Finally, the stability properties of the proposed DMPC scheme are analyzed for the case, with and without terminal constraint. We illustrate our analysis by numerical examples and compare the load on the communication network and the suboptimality for event-based communication and full communication in every iteration.