Abstract
This paper presents an efficient method for acquisition of generalized action-decision knowledge based on gaze control for virtual mobile robot. The relationships between actions and gaze control are acquired by genetic algorithm in the form of if-then rules. The extracted generalized knowledge from the acquired rules accelerate acquisition of new action-decision rules under unknown environment. The virtual mobile robot equipped with range sensors traveling from a start point to goal area is simulated. The extracted generalized knowledge from rules under some environments operated well to accelerate the learning of robot under unknown environment. The result can be interpreted as skill transition under various environments realized by means of generalization of acquired knowledge.
