Unverified50% confidenceFactExact time
The initial high-reward policy had three typical problems: some initial states intersect the ground (clipping), kneeling on both knees earns excessive reward, and the robot uses joint limits as support.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Training Whole-Body Recovery Policies for Humanoid Robots: From Reward Design to DC Motor Modeling
redditr/reinforcementlearning10/17/2026
Related Entities
Related Claims
UnverifiedThe revised reward function is designed around pelvis height, torso orientation, posture, flat foot contact, motion smoothness, and avoidance of ankle joint limits.72% similarUnverifiedProposed solutions to the four-footed grounding problem include adding foot contact rewards, adjusting torque penalty weights, incorporating symmetry constraints, and improving the anti-slip mechanism.71% similarUnverifiedAfter incorporating flat sole contact and ankle limit avoidance into the reward, the policy is more inclined to complete recovery using reasonable support methods.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1052190API
curl https://kongchang.com/api/v1/knowledge/claims/1052190MCP
get_claim(id=1052190)