Event
AME Terrain Attention Encoding Controls G1 Walking on Complex Terrain
1 reports · 1 sources
A terrain attention encoding method called AME is proposed for controlling the G1 humanoid robot to walk on complex terrain. The method encodes G1's 1.6×1.0 m elevation map into 187 terrain tokens, uses the robot's proprioceptive state as a single query for cross-attention, and extracts terrain features for use by the policy. According to the method's description, the policy can thereby walk on steps, plum blossom piles, and slopes. The related derivation involves the encoder structure, attention mechanism, and reward design.
Report timeline
- AME: Terrain-attention encoding controls G1 humanoid robot walking on complex terrain
The article introduces a terrain-attention encoding method called AME that encodes a 1.6×1.0-meter elevation map for the G1 humanoid robot into 187 terrain tokens, uses the robot's proprioceptive state as a single query for cross-attention to extract terrain features, and the policy uses these to walk on steps, plum blossom piles, and slopes. The article derives the encoder structure, attention mechanism, and reward design; the full body of the text is not provided in the materials.
Guyuehome ·