Prediction Effect on Manual Control of an Unstable System

Abstract

The study of human behavior in manual control systems is reevaluated in relation to fuzzy control. Human behavior in this problem means not only physical action but also thinking processes such as cognition, reasoning, optimization decision, and learning. The human performance of this kind is related to the interface. In this paper, we investigate experimentally the effects of an interface on the manual control of a double inverted pendulum. This system is suitable to the study of human behavior, because the stability is changed sensitively with the skills of the subjects and also with the quality of the interface. Besides the general performance of the interface, the effect of prediction is experimented in detail. Then we get some qualitative and quantitative results regarding the prediction interval, the gain constants of manipulation and display, and the prediction error. These results are well explained by a hypothesis of human skillfulness suggested in brain physiology, that is, an action-model is formed in cerebellum after practice and then human action becomes more accurate and quick because of feedforward action by this model.