Dimensions overall score 9.0

GitHub Code Pulse

No public code linked for this paper yet.

• FAME, a force-adaptive reinforcement learning framework, is proposed to condition a standing policy on a learned latent context encoding upper-body joint configuration and bimanual interaction forces.method 95%
• During training, diverse, spherically sampled 3D forces are applied on each hand to inject disturbances in simulation.method 90%
• FAME improves mean standing success to 73.84% in simulation.result 90%
• The curriculum-only baseline achieves a mean standing success of 51.40% in simulation.result 90%
• The base policy achieves a mean standing success of 29.44% in simulation.result 90%
• The learned policy is deployed on a full-scale Unitree H12 humanoid to evaluate robustness in load-interaction scenarios.result 90%
• Interaction forces are estimated from the robot dynamics and fed to the encoder, enabling online adaptation without wrist force/torque sensors.technical 90%

Competitive landscape

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

Startup potential card