基于DAMP的人形机器人复杂地形运动控制方法及系统
By constructing the DAMP reinforcement learning control framework, the problems of unstable movement and unnatural actions of humanoid robots in complex terrain are solved, and more efficient autonomous adaptive control is achieved, which is suitable for a variety of application scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SONGYAN POWER (BEIJING) TECHNOLOGY CO LTD
- Filing Date
- 2026-01-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing humanoid robots suffer from poor motion stability, insufficient naturalness of movement, inaccurate state estimation, and inconsistent control frameworks in complex terrain motion control, resulting in poor performance in variable environments.
We construct a DAMP reinforcement learning control framework that integrates a denoised world model, adversarial motion priors, dynamic reward interpolation mechanism, and Lipschitz continuity penalty. By extracting robust states through the denoised world model, generating natural human-like actions, optimizing reward allocation and policy smoothness, we can improve the robot's autonomous adaptation ability in complex terrain.
It significantly improves the motion stability and adaptability of humanoid robots in complex terrain, generates actions that conform to human behavior patterns, enhances learning efficiency and control performance, and strengthens its application potential in service, medical and entertainment fields.
Smart Images
Figure CN121848385B_ABST