Generation of evaluation function for robot force control using genetic programming

Force control is one of the most important and fundamental tasks of robot manipulators. It is known that a neuro-fuzzy control method is one of the best control methods for robot force control. Usually, the neuro-fuzzy controller is trained to minimize the error function. However, unexpected response such as overshooting or oscillation may occur when only the control error is evaluated, since the dynamics of the robot and the environment is not reflected in the evaluation function. In this paper, we propose an effective evaluation function generation method using genetic programming. The effectiveness of the proposed method was evaluated by simulation