Method and system for training motion control model, and robot
By deploying a head posture supervision model and a simulation motion control model on the head of a humanoid robot, and using image and walking state information to train the motion control model, the problems of unstable walking and high energy consumption of traditional humanoid robots in complex terrain are solved, and more efficient multi-terrain adaptation is achieved.
WO2026130285A1PCT designated stage Publication Date: 2026-06-25BEIJING ZHONGKE HUILING ROBOT TECH CO LTD
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- BEIJING ZHONGKE HUILING ROBOT TECH CO LTD
- Filing Date
- 2025-12-15
- Publication Date
- 2026-06-25
Smart Images
Figure CN2025142438_25062026_PF_FP_ABST
Abstract
The present disclosure provides a method and system for training a motion control model, and a robot. The method comprises: acquiring walking state information of a humanoid robot and image information, the image information being image information acquired when the humanoid robot performs a head movement according to a head pose control instruction during a walking process, and the image information being in one-to-one correspondence with the walking state information; determining corresponding second terrain information on the basis of the image information; and constructing a motion control model and performing model training, wherein training samples for the motion control model comprise the walking state information of the humanoid robot, the image information, and the second terrain information.
Need to check novelty before this filing date? Find Prior Art