Adaptive method of robot return function based on reinforcement learning

• Main Authors: QU HONG, FU MINGSHENG, YANG ZHIYOU, ZHANG FAN
• Format: Patent
• Language: Chinese; English
• Published: 14.03.2023
• Subjects: CONTROL OR REGULATING SYSTEMS IN GENERAL; CONTROLLING; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS ORELEMENTS; PHYSICS; REGULATING

The invention discloses an adaptive method of a robot return function based on reinforcement learning, which can learn a return value according to an interaction track of a robot and an environment, thereby guiding a reinforcement learning algorithm to optimize a control strategy, avoiding manual design intervention of a return model, and improving the reliability of the robot. The walking control of reinforcement learning in different scenes is more efficient through a self-adaptive return model. 本发明公开了基于强化学习的机器人回报函数的自适应方法，可以根据机器人与环境的交互轨迹学习到回报值，从而指导强化学习算法优化控制策略，避免了回报模型的人工设计干预，能够通过自适应的回报模型提高强化学习在不同场景下的行走控制更加高效。