Curriculum Learning Receding Horizon Energy Management for Quadruped Robot

Electrically-driven quadruped robots have become popular in the field of robotics in recent years. However, the development of reliable energy supply systems and effective energy management strategies for quadruped robots are still underexplored issues. This article proposed a curriculum learning receding horizon energy management strategy for quadruped robot powered by hybrid energy storage system. Firstly, an MPC controller is used to allocate the power between battery and supercapacitor, and the normalized objective function is designed to reduce battery capacity loss and improve robot kinematic performance. Then the online adaptive weight adjustment is implemented by using of a rule controller, which is learned from expert experience. Simulations and experiment are performed to validate the proposed method, demonstrating that the proposed method can reduce the losses in the hybrid functional system and increase its output capability.