Abstract
Many intelligent robots are controlled by the behavior rule designed by human, but the design becomes difficult for the interaction with the environment and complicated tasks. Therefore the autonomous learning of agent itself is required. As the one of them, the research to apply the concept of shaping based on the animal training to the behavior learning of robot is known recently. Shaping is a general idea that the learner is given a reinforcement signal step by step gradually and inductively forward the behavior from easy tasks to complicated tasks.
In this research, a concept of shaping is adopted to the fuzzy state division type reinforcement learning to acquire high-level behavior selection rules of the hierarchical fuzzy behavior control that was already proposed in this laboratory by reinforcement learning automatically and we propose a method which is made to learn to the robot efficiently by giving a reward appropriately by human.
