An Adaptive Training Method for Human-robot Systems Using Neural Networks

Recently, the needs of robot systems for human support such as a master-slave manipulator, a teleoperation robot and a power assist robot have been increasing. In such human-robot systems, a human operator has an initiative in executing a task, while a robot assists him or her. Therefore, the importance of a training system to improve operator's skill in controlling the robot should be recognized from a point of view of safety, since a control error of a human operator might cause a serious accident. In this paper, a new training system for human-robot systems is proposed, in which a neural network (NN) is used in order to identify the dynamic properties of the system and give an assist to the operator. The identification model used in the proposed system consists of the NN and a reference model which represents a control property of the skilled operator. This paper explains a working principle of the training method and shows the validity of the proposed method through experiments of robot control by novice operators.