NavWM: A Unified Navigation World Model for Foresight-Driven Planning

2026-06-22

AI Summary

This work addresses the limitations of conventional visual navigation—namely, myopic decision-making and mode collapse in complex environments—by proposing a unified navigation world model that jointly integrates perception, generation, and control for the first time. The approach introduces an anchor-based multimodal trajectory prediction framework coupled with a latent world token mechanism, enabling high-fidelity reasoning about future states and action prediction within a single architecture. By combining controllable visual generation with visual lookahead closed-loop planning, the model achieves coherent and robust navigation behavior. Experimental results demonstrate significant performance gains over existing methods across multiple robotic datasets, with particularly notable improvements in zero-shot navigation success rates and the quality of generated future states.

Abstract

Conventional visual navigation policies often struggle with myopic decision-making and mode collapse in complex environments. While world models offer a promising alternative, existing paradigms typically isolate perception, generation, and control, failing to capture their shared spatio-temporal dynamics. In this paper, we propose NavWM, a unified navigation world model that seamlessly integrates latent world reasoning, multimodal action prediction, and controllable visual generation. At its core, NavWM leverages latent world tokens to distill geometric and semantic priors, endowing the agent with robust structural understanding. To overcome the limitations of deterministic policies, we introduce an anchor-based multimodal trajectory forecasting framework that generates a diverse action space. This inherent diversity explicitly empowers the generative world model to act as a robust closed-loop planner, utilizing visual foresight to evaluate and select the optimal path. Extensive experiments across diverse robotics datasets demonstrate that NavWM significantly advances the state-of-the-art, delivering remarkable improvements in both high-fidelity future state generation and zero-shot navigation success.

Problem

• visual navigation
• world model
• myopic decision-making
• mode collapse
• spatio-temporal dynamics

Innovation

• unified world model
• latent world tokens
• multimodal trajectory forecasting
• visual foresight
• closed-loop planning