FlyMirage: A Fully Automated Generation Pipeline for Diverse and Scalable UAV Flight Data via Generative World Model

arXiv 2026 46.4 system, application

TLDR

FlyMirage automates generation of diverse, scalable UAV flight data using LLMs and a generative world model for aerial VLN.

Reasoning

The paper presents a novel pipeline combining LLMs and 3DGS for synthetic data generation, addressing scalability and diversity. However, it lacks real-world validation and does not evaluate on downstream tasks or compare to real datasets.

Read-first score

Read-first score 46.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 25.

Recency 6%

100

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

Citation impact 18%

82.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.82118149

Methodology quality 18%

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

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Topical relevance 29%

35.7

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

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 310.

Keyword Scores

generative world model

world model

world simulator

interactive world model

video world model

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

Using LLM as an environment designer to promote scene diversity
Generative world model that instantiates designs into high-fidelity 3D Gaussian Splatting (3DGS) scenes
Automation of scene exploration and semantic information acquisition
Dynamically feasible planner for UAV trajectory generation
Fully automated pipeline for generating large-scale, diverse, and photorealistic aerial VLN datasets

Methodology

FlyMirage is a fully automated pipeline that leverages a large language model (LLM) as an environment designer to generate diverse scene designs, which are then instantiated into high-fidelity 3D Gaussian Splatting (3DGS) scenes by a generative world model. The pipeline automates scene exploration and semantic information acquisition, and integrates a dynamically feasible planner for uncrewed aerial vehicle (UAV) trajectory generation, producing a large-scale aerial VLN dataset.

Key Results

The pipeline generates a large-scale, diverse, and photorealistic aerial VLN dataset with dynamically feasible flying trajectories, designed to support the development of next-generation embodied navigation models.

Resources

Paper