MWM: Mobile World Models for Action-Conditioned Consistent Prediction

TL;DR

MWM introduces a novel training framework for action-conditioned world models that enhances rollout consistency and inference efficiency, significantly improving visual fidelity and navigation success in embodied planning tasks.

Contribution

The paper presents a two-stage training approach with structure pretraining and ACC post-training, along with ICSD for diffusion distillation, addressing rollout inconsistency and inference efficiency in navigation models.

Findings

Improved visual fidelity and trajectory accuracy in benchmarks.

Enhanced planning success rates in real-world navigation tasks.

Increased inference efficiency with few-step diffusion distillation.

Abstract

World models enable planning in imagined future predicted space, offering a promising framework for embodied navigation. However, existing navigation world models often lack action-conditioned consistency, so visually plausible predictions can still drift under multi-step rollout and degrade planning. Moreover, efficient deployment requires few-step diffusion inference, but existing distillation methods do not explicitly preserve rollout consistency, creating a training-inference mismatch. To address these challenges, we propose MWM, a mobile world model for planning-based image-goal navigation. Specifically, we introduce a two-stage training framework that combines structure pretraining with Action-Conditioned Consistency (ACC) post-training to improve action-conditioned rollout consistency. We further introduce Inference-Consistent State Distillation (ICSD) for few-step diffusion distillation with improved rollout consistency. Our experiments on benchmark and real-world tasks demonstrate consistent gains in visual fidelity, trajectory accuracy, planning success, and inference efficiency. Code: https://github.com/AIGeeksGroup/MWM. Website: https://aigeeksgroup.github.io/MWM.