The Application of a Theoretical Learning Model to a Remote Handling Control System

摘要

The report introduces the concept of adaptive aiding as a means of improving performance in remote handling, and describes the development, realization and experimental evaluation of a system based on this concept. The system incorporates an automaton that is able to supplement the operator's control of a powered manipulator arm as well as act as a parallel controller. The operator positions the manipulator in three dimensions by means of joystick control; a foot-pedal opens and closes the manipulator claw. The behavior of the automaton is established through a process of learning -- observing the human operator's control actions relative to the environment and the manipulator output, and developing a decisionmaking policy. Initially, the manipulator is totally controlled by the operator, while the automaton acts as a passive observer. As operation continues, behavior patterns are acquired by the automaton, and gradually it is transformed from a passive observer to an active controller. (Author)