Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• SCDP enables humanoid locomotion using only onboard sensors, eliminating the need for explicit state estimation.method 95%
• SCDP achieves near-perfect success (99-100%) on velocity-commanded locomotion tasks in simulation.result 90%
• SCDP achieves 93% tracking success on the AMASS test set for motion reference tracking tasks in simulation.result 90%
• SCDP performs comparably to privileged baselines that use full-body states, while using only onboard sensors.result 85%
• The trained SCDP policy was successfully deployed on a real G1 humanoid robot at 50 Hz.result 95%
• The deployed policy demonstrates robust real robot locomotion without external sensing or state estimation.result 85%
• SCDP's mixed-observation training enforces the model to infer motion dynamics under partial observability.technical 90%
• The method incorporates restricted denoising, context distribution alignment, and context-aware attention masking to prevent train-deploy mismatch.technical 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card