RecoverFormer: End-to-End Contact-Aware Recovery for Humanoid Robots

TL;DR

RECOVERFORMER is an end-to-end humanoid recovery policy using a causal transformer that switches among recovery behaviors and predicts environmental contact points, demonstrating robust zero-shot transfer and generalization.

Contribution

The paper introduces RECOVERFORMER, a novel transformer-based architecture with latent recovery modes and contact affordance heads for contact-aware humanoid recovery.

Findings

Achieves 100% recovery success in walled environments under various pushes.

Transfers zero-shot to unseen environments with high success rates.

Maintains high performance under model mismatch, latency, and friction perturbations.

Abstract

Humanoid robots operating in unstructured environments must recover from unexpected disturbances-a capability that remains challenging for end-to-end control policies. We present RECOVERFORMER, a fully end-to-end humanoid recovery policy that learns when and how to switch among recovery behaviors-including compensatory stepping, hand-environment contact, and center-of-mass reshaping-while maintaining robust performance under model mismatch. The architecture combines a causal transformer over a 50-step observation history with two novel heads: a latent recovery mode that enables smooth transitions among distinct recovery strategies, and a contact affordance head that predicts which environmental surfaces (walls, railings, table edges) are beneficial for stabilization. We evaluate RECOVERFORMER on the Unitree G1 humanoid in MuJoCo. Trained only on open floor, RECOVERFORMER transfers zero shot to walled environments, achieving 100% recovery success across 100-300 N pushes and across wall distances from 0.25-1.4m. Under zero-shot dynamics mismatch, RECOVERFORMER reaches 75.5% at plus +25% mass, 89% under 30 ms latency, 91.5% at low friction, and 99% under compound friction, latency and mass perturbation. The learned latent modes specialize across force regimes without mode-level supervision, validated by t-SNE analysis of 300 episodes. Taken together, these results show that a single end-to-end policy can deliver multi-modal, contact aware humanoid recovery that generalizes across perturbation magnitude, contact geometry, and dynamics shift.