A behavior learning/operating module for mobile robots

Abstract

Robot behavior learning is an emerging research topic in robotics. By incorporating learning capability to robots, engineers are not required to hard-code appropriate actions under every possible situation. Actually, this is an impossible task. In this paper, an architecture of behavior learning/operating module (BLOM) for a robot system is proposed. In the BLOM architecture, several categories of situations and actions are formed and mappings among the situation and action categories are established. A Knight Tournament (KT) strategy is proposed for adaptive categorization of situation and action patterns in learning. A computer simulation on learning a robot behavior is also presented.