Neural-Network-Based Adaptive Controller for Uncertainty Compensation of Robot Manipulators

Abstract

The uncertainties (parametric and/or structural) of the robot manipulator make the controller design problem difficult. In this paper, the controller consists of a model-based control law (an ineritia-rclalcd method) and a neural-network compensator. Neural network is used to adaptively compensate for the uncertainties (parametric and/or structural). The closed-loop system is shown to be stable in the Lyapunov sense, A novel adaptive learning algorithm for tuning the neural-network weights is derived. The error signal for tuning the neural network can be easily obtained from the controller design without using any model knowledge of the robot manipulator in this adaptive learning algorithm. Simulation results show that fast learning for neural network is possible, and the closed-loop system can achieve superior tracking performance.