Disturbance-Aware Adaptive Compensation in Hybrid Force-Position Locomotion Policy for Legged Robots
Yang Zhang, Zhanxiang Cao, Buqing Nie, Yangqing Fu, Haoyang Li, Zheng Zhang, Yizhi Chen, Xiaokang Yang, Yue Gao
AI summary
Problem
Existing reinforcement learning locomotion policies rely solely on target joint positions, causing delayed and indirect responses to real-world payload variations and external disturbances that degrade stability.
Approach
The framework outputs both target joint positions and feedforward torques, paired with a neural disturbance observer that estimates external forces to generate real-time torque compensation.
Key results
- Hybrid force-position action space improves joint tracking accuracy under contact forces
- Neural disturbance observer explicitly estimates external forces without precise dynamic models
- Adaptive compensation policy actively counteracts unknown disturbances in the torque space
- Zero-shot real-robot deployment on Unitree Go2 demonstrates superior payload capacity and disturbance rejection
Why it matters
Enables legged robots to operate reliably in unstructured, dynamic environments where payloads and terrain forces vary unpredictably.
Abstract
Reinforcement Learning (RL)-based methods have significantly improved the locomotion performance of legged robots. However, these motion policies face significant chal- lenges when deployed in the real world. Robots operating in uncertain environments struggle to adapt to payload variations and external disturbances, resulting in severe degradation of motion performance. In this work, we propose a novel Hybrid Force-Position Locomotion Policy (HFPLP) learning framework, where the action space of the policy is defined as a combination of target joint positions and feedforward torques, enabling the robot to respond rapidly to payload variations and external disturbances. In addition, the proposed Disturbance-Aware Adaptive Compensation (DAAC) provides compensation actions in the torque space based on external disturbance estimation, enhancing the robot’s adaptability to dynamic environmental changes. We validate our approach in both simulation and real-world deployment, demonstrating that it outperforms existing methods in carrying payloads and resisting disturbances.