強化学習に基づく4脚走行ロボットのバウンド歩容獲得

～状態重視型と行動重視型の提案と検討～

抄録

Robots have been rapidly developing with advances in technology. No longer limited to factory use, they can now be found, operating in homes and cities. However, many of these robots are fundamentally set up to repeat the same operation, and have difficulty adapting to environmental changes and unexpected situations. Reinforcement learning can be used to solve this problem. Reinforcement learning is a machine learning method. The technique determines an optimal behavior pattern with respect to the target depending on the circumstances and environment by repeating trial and error. Using this method, robots will attain or change an optimal behavior pattern based on the changes in environment and situation. In this study, we applied reinforcement learning to a robot modeled on a four-legged animal and examined whether the robot could attain running behavior (bound gait). At the conclusion, our robot was able to attain stable running motion, and it shew much improvement in velocity by changing the reward.