Particle Swarm Optimization for use with Distributed Model Predictive Control

In Distributed Model Predictive Control (DMPC) the goals of a centralized Model Predictive Control (MPC) system are achieved using communication between agents, in order to coordinate individual agents´ responses. This can be achieved using an augmented Lagrangian formulation, which is then distributed between agents using Block Coordinate Descent. Minima in these problems are typically found using deterministic optimization techniques. This paper examines the possible application of Particle Swarm Optimization for this purpose. This stochastic optimization technique is shown to perform similarily to the deterministic techniques in terms of arriving at suitable control inputs but with increased optimization time.