Control systems in robots

Abstract

A general framework is presented for treating the production of behavior in a system that interacts purposively with its environment. Such a system is viewed as having two parts: a body, which forms the interface with the outside world, and a control. Communications between these two parts constitute the moves in the system's game against the environment. The control embodies the rules that specify the system's set of tasks, as well as the strategy used in the reduced game that results from the use of the rules. Three classes of strategies are defined: local, problem-solving, and decision-making. Decision-making strategies, applied in the reduced game with the policy of maximizing utility, have desirable features. The simulation of PERCY, a nest-building insect, illustrates how such a strategy avoids difficulties that obstruct a problem-solving approach.