The ACO-NM algorithm for controller tuning for an inverted cart-pendulum

Abstract

The control of a cart-pendulum system consists to swing the pendulum up and to maintain it in its upright position. In many publications, one controller is used for the swing-up action and one for the stabilization mode. Both controllers are usually tuned separately and frequently by trial-and-error, as is the handling of system constraints. This paper proposes the simplified Ant Colony Optimization algorithm combined to a constrained Nelder-Mead method (ACO-NM) for cart-pendulum control structure tuning. A new cost function is proposed to manage cart-rail length. Simulation and experimental results confirm the efficiency of the ACO-NM algorithm to optimize the control structure for four cart-pendulum parameter sets.