2P1-F05 Incremental Behavior Acquisition Based on Reliability of Observed Behavior Recognition
Tomoki NISHI, Yasutake Takahashi, Minoru Asada · The Proceedings of JSME annual Conference on Robotics and Mechatronics (Robomec) · 2007
We propose a novel approach for acquisition and development behaviors through observation of behaviors in multi-agent environment. Observed behaviors of others gives fruitful hints to find a new situation, a new behavior for the situation, necessary information for the behavior acquisition. RoboCup scenario gives us a good test-bed multi-agent environment where a learner can observe behaviors of others during practices or games. It is more realistic, practical, and efficient to take advantages of observation of skilled players than to discover new skills and necessary information only though the interaction of robot and environment. The robot automatically detects state variables and a goal of the behavior through the observation based on mutual information. Reinforcement learning method is applied to acquire the discovered behavior suited to the robot. Experiments under RoboCup Middle Size scenario shows the validity of the proposed method.