Development of Bipedal Locomotion Controller with Model-based Reinforcement Learning

基于模型的强化学习双足运动控制器的开发

• 批准号: 22KJ2291
• 负责人: Kuo ChengーYu
• 金额: $ 1.09万
• 依托单位: Nara Institute of Science and Technology
• 依托单位国家: 日本
• 项目类别: Grant-in-Aid for JSPS Fellows
• 财政年份: 2023
• 资助国家: 日本
• 起止时间: 2023-03-08 至 2024-03-31
• 项目状态: 已结题
• 来源: https://kaken.nii.ac.jp/en/grant/KAKENHI-PROJECT-22KJ2291/
• 关键词: モデルベース強化学習; 弾性ロボット; 二脚ロボット; 確率手法

During this year, we utilized model-based reinforcement learning to capture the dynamics of a spring-loaded biped robot. This is significant because the robot's compliancy makes modeling analytic dynamics difficult. Using the learned dynamics, we successfully performed hopping tasks with real-time planning in a simulation environment. These results demonstrate the effectiveness of our approach. Additionally, we have begun hardware implementation, achieving a simple walking task. The performance is expected to significantly improve in the following year.

今年，我们利用基于模型的强化学习来捕捉弹簧加载的双足机器人的动态。这很重要，因为机器人的顺应性使得分析动力学建模变得困难。利用学习到的动力学，我们在模拟环境中通过实时规划成功执行了跳跃任务。这些结果证明了我们方法的有效性。此外，我们已经开始硬件实现，实现简单的步行任务。预计来年业绩将显着改善。