Awesome World Model Hub Papers · Datasets · Projects
← Back to papers

An Efficient and Multi-Modal Navigation System with One-Step World Model

arXiv 26.1 2026 64.3 method, system

TLDR

A lightweight one-step world model with 3D U-Net for efficient multi-modal navigation, validated in simulation and real-world.

Reasoning

The paper proposes a novel one-step generation world model to overcome latency issues in navigation, with strong empirical validation in both simulation and real-world closed-loop experiments. However, the abstract lacks detailed comparison of limitations or ablation studies, and the reliance on anchor-based initialization may limit generality.

Read-first score

Read-first score 64.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 52.

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 42%
74.3

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Methodology quality 25%
70

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,experiment,result

Reproducibility 25%
30

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 267.

Keyword Scores

world model
10
world dynamics prediction
9
generative world model
8
interactive world model
7
video world model
7
world simulator
6
model-based reinforcement learning world model
5

Deep Analysis

Innovations

  • One-step generation paradigm for navigation world model, replacing multi-step diffusion and autoregressive generation
  • 3D U-Net backbone with efficient spatial-temporal attention for lightweight and fast inference
  • Anchor-based initialization in optimization-based planning for multi-modal goal navigation

Methodology

The paper proposes a lightweight navigation world model using a one-step generation paradigm with a 3D U-Net backbone and efficient spatial-temporal attention to reduce inference latency. This model is integrated into an optimization-based planning framework with anchor-based initialization to handle multi-modal goal navigation tasks. The system is evaluated through closed-loop experiments in both simulation and real-world environments.

Key Results

The proposed system demonstrates superior efficiency and robustness compared to state-of-the-art baselines in extensive closed-loop experiments in simulation and real-world environments.

Tags