FUZZY SUPERVISION OF ROBOTIC EXCAVATOR RESEARCH ON NEURAL NETWORK CONTROLLERS

A method of rule-based self-learning neural-network controller applied to the robotic excavator is presented. According to the real-time control result, the controller takes advantage of the method of multiple-steps learning and fuzzy supervision to correct the teacher signals of neural-network, therefore, the control algorithm becomes simple, the reality of calculating is improved, and the learning speed is increased. This method is tested and verified by experiments.