An Event-Triggered Dual Averaging Algorithm for Distributed Model Predictive Control

This article proposes an iterative distributed model predictive control (DMPC) approach for discrete-time linear systems with decoupled and coupled constraints over undirected communication networks. To address the coupled constraint, we transform a DMPC optimization problem into its dual problem. Then, a novel event-triggered distributed dual averaging (ETDDA) algorithm is designed to solve this dual problem, in which event-triggered mechanisms can tremendously save communication resources. We prove that the convergence of the proposed ETDDA algorithm can be guaranteed as long as the step-size satisfies the specifically designed condition and the triggering threshold bound is square summable. Moreover, we propose a distributed stopping criterion to terminate our ETDDA algorithm when the output solution meets the specified optimality and the coupled constraint is satisfied, thereby avoiding infinite iterations and saving computational resources. The recursive feasibility and the closed-loop stability under the proposed iterative DMPC approach are strictly proved. Finally, a simulation study shows that our approach performs better compared to its competitors.