Adaptive iterative learning control of robotic system based on particle swarm optimization

Abstract

In this paper, an adaptive iterative learning control algorithm for robotic manipulators is proposed. A simplified robot manipulator model with 3 degrees of freedom is used as control object for verification purposes. The mathematical model is obtained via Rodriguez approach for modeling differential equations of motion for multi-body systems. The model itself is a simple open-chain kinematic structure. The proposed control system design consists of two layers of controllers. In the inner loop, feedback linearization is applied to deal with the model nonlinearities. Post feedback linearization advanced iterative learning control (ILC) algorithm of sign-D (signum-Derivative) type is introduced as feed-forward compensation with classical PD (Proportional-Derivative) controller in feedback closed loop. A particle swarm optimization (PSO) algorithm is used to optimize ILC gain parameters while gains for PD controller are set by trial and error. Suitable cost function based on position error is chosen for PSO algorithm in order to ensure convergence. Numerical simulation is carried out in two cases – case with constant learning gains and case with PSO optimized learning gains. It is observed that the proposed control law converges to some steady-state error value in both cases.