Aerial World Model for Long-horizon Visual Generation and Navigation in 3D Space

arXiv 25.12 2025 61.3 method, application

TLDR

Proposes ANWM, an aerial navigation world model that predicts future visual observations for long-horizon UAV navigation using a physics-inspired Future Frame Projection module.

Reasoning

The paper introduces a novel physics-inspired module (FFP) to improve long-distance visual forecasting and navigation success, which is a strength. However, it is limited to 4-DoF trajectories and does not explicitly validate on real-world data, weakening its generalizability claim.

Read-first score

Read-first score 61.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 53.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Topical relevance 42%

75.7

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%

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

Reproducibility 25%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 373.

Keyword Scores

world model

generative world model

video world model

world dynamics prediction

interactive world model

world simulator

model-based reinforcement learning world model

Deep Analysis

Innovations

Proposes ANWM, an aerial navigation world model that predicts future visual observations conditioned on past frames and actions for high-level semantic planning.
Introduces a physics-inspired Future Frame Projection (FFP) module that projects past frames into future viewpoints to provide coarse geometric priors, reducing representational uncertainty in long-distance visual generation.

Methodology

ANWM is trained on 4-DoF UAV trajectories and uses the Future Frame Projection (FFP) module to project past frames into future viewpoints, providing geometric priors. The model enables agents to rank candidate trajectories by semantic plausibility and navigational utility, bridging low-level control with high-level semantics.

Key Results

ANWM significantly outperforms existing world models in long-distance visual forecasting and improves UAV navigation success rates in large-scale environments.

Resources

Paper