LUCID: Latent-Skill Unified Control via Imagined Dynamics for Long-Horizon Humanoid Loco-Manipulation

2026-08-07

AI Summary

Existing approaches to long-horizon mobile manipulation tasks for humanoid robots struggle to flexibly compose diverse whole-body skills and achieve robust high-level decision-making, often relying on scripted planners or task-specific policies. This work proposes LUCID, a novel framework that, for the first time, integrates reusable structured implicit skill policies with imagination-based trajectory planning. By jointly optimizing a high-level policy and a macro-dynamics world model through hierarchical model-based reinforcement learning, LUCID eliminates the need for task-specific engineering. The method substantially improves both overall success and partial completion rates on multi-object rearrangement tasks, outperforming current baselines without requiring handcrafted task designs.

Abstract

Long-horizon humanoid loco-manipulation requires composing versatile whole-body skills and reliable high-level decision making. Existing methods often coordinate pretrained skills with scripted planners, finite-state machines or task-specific model-free policies, restricting their ability to handle complex task sequences. To address this limitation, we propose \textbf{LUCID}, a hierarchical model-based reinforcement learning framework that plans over reusable skills through imagined rollouts of a learned dynamics model. LUCID first trains a structured latent-conditioned low-level policy via adversarial imitation and then freezes it while jointly learning a high-level policy and macro-dynamics world model. The world model predicts the temporally extended state transitions induced by latent decisions, enabling high-level policy optimization through imagined rollouts. We evaluate our framework across various simulated multi-object rearrangement scenarios. Experimental results show that LUCID improves the full-task success and partial-completion rates compared to prior baseline methods, demonstrating its effectiveness in complex sequential loco-manipulation tasks.

Problem

• long-horizon
• humanoid
• loco-manipulation
• skill composition
• high-level decision making

Innovation

• hierarchical reinforcement learning
• learned dynamics model
• latent-conditioned policy
• imagined rollouts
• loco-manipulation