Driving · 2026-10-15
Contact-UAN: Unsupervised Actuator Networks as a Residual World Model for Humanoid Control
Nolan Fey, Khai Nguyen, Yash Prabhu, Gabriel B. Margolis, Pulkit Agrawal
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
Contact-UAN is an unsupervised residual world model for humanoid control that preserves a simulator’s physics priors while correcting actuator dynamics to match hardware rollouts. It uses contact-consistent replay by inferring simulator-side stance-foot pose from onboard proprioception so residual learning corrects actuator discrepancies rather than contact/reset artifacts. A UAN trained on 6.5 minutes of real-world walking transfers to running, force control/impedance, and robust jumping.
Why it matters
The authors address sim-to-real failures in humanoid control where force interactions learned in simulation do not transfer to hardware due to actuator sim-to-real gaps. Contact-UAN targets this by learning an actuator-channel residual and enforcing contact-consistent state initialization for replay-based residual training from onboard proprioception, enabling improved force/torque behavior and robust impacts.
Method
- Contact-consistent replay initialization: infer simulator-side stance-foot pose (including small penetration/tilt) and base velocity from short proprioceptive history to match hardware contact wrenches.
- Train a residual world model (UAN) that outputs joint-position target corrections added to a base target, preserving simulator rigid-body dynamics while modulating only the actuator command channel.
- Train downstream task policies in the UAN-augmented simulator (with robot-specific domain randomization removed) and deploy only the task policy on hardware.
Limitation
The method requires useful rigid-body and contact dynamics, but not an accurately identified actuator model.
Abstract (from arXiv)
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. This failure can arise from actuator sim-to-real gaps because forces depend on actuator torques. We introduce Contact-UAN, a residual world model that preserves a simulator's physics priors while correcting actuator dynamics to align simulated rollouts with recorded hardware rollouts. 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. Using only onboard proprioception, Contact-UAN infers the simulator-side stance-foot pose needed to initialize replay with consistent contact forces. We show that a UAN trained on 6.5 minutes of real-world walking data generalizes to tasks that benefit from accurate force and impulse regulation and produces policies that transfer seamlessly to the real world, yielding 20% lower running torque, 43% lower force-tracking error, extreme compliance, accurate impedance control, and dynamic, robust jumps.
Related papers
- Being-M0.7: A Latent World-Action Model for Humanoid Robots
- LocoWM: High-Precision Locomotion through World-Model-Guided Residual Adaptation
- VioLA: Learning Generalist Humanoid Control Policies from Human Data
- BeJEPA: A Goal-Conditioned Humanoid Behavior World Model via Unsupervised RL
- Action-Aligned Latent Planning for Versatile Humanoid Loco-Manipulation
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-13 · Indexed by Pub-AI: 2026-10-15 UTC
Contact-UAN: Unsupervised Actuator Networks as a Residual World Model for Humanoid Control
Preprint posted: 2026-10-13 · Original record ↗
- arxiv
- 2610.18991
Explicit source versions (1)
- v1 · 2026-10-13
- Preprint available
Source date: 2026-10-13 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
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. This failure can arise from actuator sim-to-real gaps because forces depend on actuator torques. We introduce Contact-UAN, a residual world model that preserves a simulator's physics priors while correcting actuator dynamics to align simulated rollouts with recorded hardware rollouts. 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. Using only onboard proprioception, Contact-UAN infers the simulator-side stance-foot pose needed to initialize replay with consistent contact forces. We show that a UAN trained on 6.5 minutes of real-world walking data generalizes to tasks that benefit from accurate force and impulse regulation and produces policies that transfer seamlessly to the real world, yielding 20% lower running torque, 43% lower force-tracking error, extreme compliance, accurate impedance control, and dynamic, robust jumps.