Optimal Tracking Control of DC Motors with Partially Unknown Dynamics

Abstract

In this article, the nearly optimal tracking control scheme is established for dc motors with partially uncertain dy-namics, including unknown drift dynamic and known control direction. The proposed optimal tracking control approach adopts dynamic inversion concept to acquire the feedforward compensation and then adopts reinforcement learning techniques to acquire the optimal feedback. The former derived by feedforward neural network is leveraged to estimate the unknown drift dynamic, and the latter is generated by direct heuristic dynamic programming (dHDP) to solve a Hamilton-Jacobi-Bellman (HJB) equation. Eventually, theoretical analysis demonstrates that the closed-loop system signals are all bounded and the developed control scheme can achieve the optimal control input with a small bounded error. Simulation examples are implemented to validate the feasibility of the designed method.