SplitAdapter: Load-Aware Humanoid Loco-Manipulation via Factorized Adaptation

2026-06-02

AI Summary

This work addresses the instability in whole-body control of humanoid robots when manipulating objects of varying masses and performing pick-and-place tasks at different heights, a challenge exacerbated by load-induced dynamics mismatch—particularly during sim-to-real transfer. To tackle this, the authors propose SplitAdapter, an architecture that freezes a pretrained policy and introduces decoupled context encoders for load and dynamics awareness. By integrating a factorized world model, gradient reversal layer (GRL)-based cross-adversarial regularization, and a hierarchical FiLM mechanism, the approach enables factorized adaptation without relying on a monolithic latent representation, which often degrades under heavy loads. Evaluated across tasks involving 2/4/6 kg payloads and 0/30/60 cm target heights, SplitAdapter achieves substantially higher success rates than baseline methods, with especially pronounced gains in heavy-load scenarios.

Abstract

Humanoid loco-manipulation requires stable whole-body control under varying object masses and pickup/placement heights. This becomes particularly challenging in sim-to-real transfer, where object-induced load variation and robot-side dynamics mismatch interact during physical contact. Existing history-based adapters often compress these factors into a single latent representation, which can weaken robustness under heavy-load manipulation. We propose \textbf{SplitAdapter: Load-Aware Humanoid Loco-Manipulation via Factorized Adaptation}, which freezes a pretrained box manipulation policy and extends it with object/load and dynamics-aware context encoders trained with split world-model objectives, GRL-based cross-adversarial regularization, and hierarchical Feature-wise Linear Modulation (FiLM). In sim-to-sim experiments and real-world deployment, SplitAdapter improves Full-task success over the base policy and world-model FiLM baselines across object masses of $2$, $4$, and $6$ kg and pickup/placement heights of $0$, $30$, and $60$ cm, with the largest improvements under heavy-load conditions.

Problem

• humanoid loco-manipulation
• sim-to-real transfer
• load variation
• dynamics mismatch
• whole-body control

Innovation

• Factorized Adaptation
• Load-Aware Control
• Sim-to-Real Transfer
• Hierarchical FiLM
• Cross-Adversarial Regularization