A metaheuristic particle swarm optimization approach to nonlinear model predictive control

Abstract

This paper commences with a short review on optimal control for nonlinear systems, emphasizing the Model

Predictive approach for this purpose. It then describes the Particle Swarm Optimization algorithm and how it could be applied

to nonlinear Model Predictive Control. On the basis of these

principles, two novel control approaches are proposed and anal- ysed. One is based on optimization of a numerically linearized

perturbation model, whilst the other avoids the linearization step altogether. The controllers are evaluated by simulation of an inverted pendulum on a cart system. The results are compared with a numerical linearization technique exploiting conventional

convex optimization methods instead of Particle Swarm Opti- mization. In both approaches, the proposed Swarm Optimization

controllers exhibit superior performance. The methodology is then extended to input constrained nonlinear systems, offering a promising new paradigm for nonlinear optimal control design.