Heuristic Dynamic Programming in Wheeled Mobile Robot Control

In this paper a discrete algorithm for tracking control of a two-wheeled mobile robot is presented. The basis of the control algorithm is an Adaptive Critic Design in Heuristic Dynamic Programming (HDP) configuration. HDP is a model-based discrete reinforcement learning algorithm. In proposed control algorithm Actor - Critic structure is supplied by a PD controller and a supervisory element. The algorithm does not require preliminary learning, works on-line and uses a dynamics model of the mobile robot for a state prediction in Actor - Critic structure. The performance of control algorithm was tested by experiments on the mobile robot Pioneer-2DX.