Particle swarm optimization for nonlinear model predictive control

Abstract

The paper proposes two Nonlinear Model Predictive Control schemes that uncover a synergistic relationship between on-line receding horizon style computation and Particle Swarm Optimization, thus benefiting from both the performance advantages of on-line computation and the desirable properties of Particle Swarm Optimization. After developing these techniques for the unconstrained nonlinear optimal control problem, the entire design methodology is illustrated by a simulated inverted pendulum on a cart, and compared with a particular numerical linearization technique exploiting conventional convex optimization methods. This is then extended to input constrained nonlinear systems, offering a promising new paradigm for nonlinear optimal control design.