A practical fuzzy controller with Q-learning approach for the path tracking of a walking-aid robot

Abstract

This study tackles the path tracking problem of a prototype walking-aid (WAid) robot which features the human-robot interactive navigation. A practical fuzzy controller is proposed for the path tracking control under reinforcement learning ability. The inputs to the designed fuzzy controller are the error distance and the error angle between the current and the desired position and orientation, respectively. The controller outputs are the voltages applied to the left- and right-wheel motors. A heuristic fuzzy control with the Sugeno-type rules is then designed based on a model-free approach. The consequent part of each fuzzy control rule is designed with the aid of Q-learning approach. The design approach of the controller is presented in detail, and effectiveness of the controller is demonstrated by hardware implementation and experimental results under human-robot interaction environment. The results also show that the proposed path tracking control methods can be easily applied in various wheeled mobile robots.