H∞ Optimal Tracking Control for Robot Manipulators with Input Constraint and Disturbances

Abstract

This paper introduces an H∞ optimal tracking controller for robot manipulators with saturation control torque, and disturbances based on a reinforcement learning (RL) method. The robot manipulators dynamics are transformed into a strict-feedback nonlinear system with input constraint, and disturbances. The feedforward control inputs are designed to transform a position tracking control problem into an H∞ optimal control problem. The constrained Hamilton–Jacobi–Isaac (HJI) equation is built, which is solved by the online RL algorithm using only a single neural network (NN) to reduce computational cost. Then the optimal control law with the input constraint and the worst disturbance law are determined. The concurrent learning (CL) technique is used to relax the demand for the persistence of excitation (PE) condition when updating the critic NN weights to global optimal values. The control performance is given in the comparative simulation results, showing the effectiveness of the proposed method.