Learning Feed-Forward Control for a Two-Link Rigid Robot Arm

Abstract

This paper introduces a control structure which consists of a Proportional Derivative (PD) controller and a Neural Network (NN)-based Learning Feed-Forward Controller (LFFC) to a Two-Link Rigid Robot Arm. An online B-spline neural network is used because of its local weight-updating characteristic, which has the advantages of fast convergence speed and low computation complexity. The torque applied to each link is defined using the EulerLagrange equation. The controller design takes into account the troubles caused by inertial loading, coupling reaction forces between joints, and gravity loading effects. This control structure can be directly applied to different robots within the same class with different lengths and masses. Simulation results are presented to demonstrate the robustness of our proposed controller under serve changes of the system parameters. 