AstraNav-World: World Model for Foresight Control and Consistency

2025-12-25

AI Summary

To address the challenges of long-horizon foresight and physical consistency in embodied navigation within open, dynamic environments, this paper proposes an end-to-end world model that jointly models future visual states and action sequences, enabling simultaneous visual prediction and action planning within a probabilistic framework. Our method introduces a novel bidirectional constraint training mechanism, ensuring that visual predictions are executable and that actions are grounded in task-relevant futures—thereby departing from the conventional decoupled “imagine-then-plan” paradigm. Integrating diffusion-based video generation with vision-language policy networks, we jointly optimize action-conditioned multi-step visual forecasting and vision-conditioned trajectory generation. Extensive evaluation across multiple embodied navigation benchmarks demonstrates significant improvements in trajectory accuracy and success rates. Moreover, zero-shot transfer to real-world environments yields strong performance, validating the model’s robust generalization in spatial understanding and navigation dynamics.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Embodied navigation in open, dynamic environments demands accurate foresight of how the world will evolve and how actions will unfold over time. We propose AstraNav-World, an end-to-end world model that jointly reasons about future visual states and action sequences within a unified probabilistic framework. Our framework integrates a diffusion-based video generator with a vision-language policy, enabling synchronized rollouts where predicted scenes and planned actions are updated simultaneously. Training optimizes two complementary objectives: generating action-conditioned multi-step visual predictions and deriving trajectories conditioned on those predicted visuals. This bidirectional constraint makes visual predictions executable and keeps decisions grounded in physically consistent, task-relevant futures, mitigating cumulative errors common in decoupled "envision-then-plan" pipelines. Experiments across diverse embodied navigation benchmarks show improved trajectory accuracy and higher success rates. Ablations confirm the necessity of tight vision-action coupling and unified training, with either branch removal degrading both prediction quality and policy reliability. In real-world testing, AstraNav-World demonstrated exceptional zero-shot capabilities, adapting to previously unseen scenarios without any real-world fine-tuning. These results suggest that AstraNav-World captures transferable spatial understanding and planning-relevant navigation dynamics, rather than merely overfitting to simulation-specific data distribution. Overall, by unifying foresight vision and control within a single generative model, we move closer to reliable, interpretable, and general-purpose embodied agents that operate robustly in open-ended real-world settings.

Problem

• Develops world model for foresight in dynamic navigation environments
• Unifies visual prediction and action planning to reduce errors
• Enables zero-shot adaptation to unseen real-world scenarios

Innovation

• Unified probabilistic framework for joint visual-action reasoning
• Diffusion-based video generator integrated with vision-language policy
• Bidirectional training optimizing visual predictions and action trajectories