Contact-UAN: Unsupervised Actuator Networks as a Residual World Model for Humanoid Control

What happened

arXiv:2610.18991v1 Announce Type: new Abstract: Humanoid policies trained in simulation can transfer qualitatively to hardware while failing to transfer the force interactions learned during training, compromising torque efficiency, load regulation, and robustness to impacts. For humanoid motion with contact, this requires contact-consistent replay: if hardware playback in simulation begins with misaligned contact wrenches, the residual model learns to correct reset artifacts rather than actuator (the part of a machine that produces movement) discrepancies.

This failure can arise from actuator sim-to-real gaps because forces depend on actuator torques. Using only onboard proprioception, Contact-UAN infers the simulator-side stance-foot pose needed to initialize replay with consistent contact forces.

Key facts

• For humanoid motion with contact, this — requires: contact-consistent replay: if hardware playback in simulation begins with misaligned contact wrenches, the residual model learns to correct reset artifacts rather than actuator discrepancies