LessMimic: Long-Horizon Humanoid Interaction with Unified Distance Field Representations

TL;DR

LessMimic introduces a unified, reference-free approach for long-horizon humanoid interaction using Distance Fields, enabling geometric generalization, skill composition, and vision-only deployment without motion capture.

Contribution

It presents a novel Distance Field-based policy framework that generalizes across object scales, composes multiple skills, and transfers seamlessly to vision-only systems.

Findings

Achieves 80-100% success on PickUp and SitStand tasks across various object sizes.

Maintains viability up to 40 sequentially composed tasks.

Outperforms baselines sharply degrading with object scale variations.

Abstract

Humanoid robots that autonomously interact with physical environments over extended horizons represent a central goal of embodied intelligence. Existing approaches rely on reference motions or task-specific rewards, tightly coupling policies to particular object geometries and precluding multi-skill generalization within a single framework. A unified interaction representation enabling reference-free inference, geometric generalization, and long-horizon skill composition within one policy remains an open challenge. Here we show that Distance Field (DF) provides such a representation: LessMimic conditions a single whole-body policy on DF-derived geometric cues--surface distances, gradients, and velocity decompositions--removing the need for motion references, with interaction latents encoded via a Variational Auto-Encoder (VAE) and post-trained using Adversarial Interaction Priors (AIP) under Reinforcement Learning (RL). Through DAgger-style distillation that aligns DF latents with egocentric depth features, LessMimic further transfers seamlessly to vision-only deployment without motion capture (MoCap) infrastructure. A single LessMimic policy achieves 80--100% success across object scales from 0.4x to 1.6x on PickUp and SitStand where baselines degrade sharply, attains 62.1% success on 5 task instances trajectories, and remains viable up to 40 sequentially composed tasks. By grounding interaction in local geometry rather than demonstrations, LessMimic offers a scalable path toward humanoid robots that generalize, compose skills, and recover from failures in unstructured environments.