Combining a rule-based expert system and machine learning in a simulated mobile robot control system

This paper presents the design of a novel, dynamic, rule-based expert system used for simulated robot control. Machine teaming techniques are used to enable the robot to collect training examples by autonomously exploring an unknown environment. As the interaction between the robot and the environment is dynamic, sequences of perceptions and sequences of events should determine future actions. Thus, temporal information is encoded into the rule-base. It is shown that the robot is able to incrementally develop a rule-base that enables it to successfully navigate through its environment.