Flexible Joint Manipulator Control Based on Adaptive Dynamic Programming

摘要

The authors present an adaptive dynamic programming-based (ADP-based) scheme to solve the tracking control problems of flexible joint manipulator, which can be transformed to solve the Hamilton-Jacobi-Bellman (HJB) equation. For the nominal error system, a critic neural network is established to approximate the value function and an online learning algorithm is proposed to update the weight of the critic neural network. The weight update of the critic neural network occurs at the same time as the system is running. The dynamics of the weight estimation error and the tracking error are proven to be uniformly ultimately bounded via Lyapunov method. Finally, a numerical example is given to illustrate the performance of the present method.