AirNav: A Large-Scale UAV Vision-and-Language Navigation Dataset with Natural and Diverse Instructions

TL;DR

AirNav is a large-scale UAV navigation dataset with natural instructions, enabling realistic training and evaluation of UAV vision-and-language models, and demonstrating state-of-the-art performance and transferability.

Contribution

The paper introduces AirNav, a comprehensive UAV navigation dataset with natural instructions, and proposes AirVLN-R1, a model achieving state-of-the-art results with real-world transferability.

Findings

AirVLN-R1 achieves 51.82% success rate on test-unseen split.

The dataset includes 137K navigation samples with natural instructions.

Real-world UAV experiments suggest promising sim-to-real transfer.

Abstract

Existing UAV vision-and-language navigation (VLN) benchmarks rarely provide realistic aerial scenes, natural process-level instructions, and sufficient scale simultaneously, making it difficult to systematically train and evaluate UAV VLN agents under realistic settings. To address this, we propose \textbf{AirNav}, a large-scale benchmark built on real urban aerial data, comprising 137K navigation samples with natural and diverse instructions generated via a human--LLM collaborative pipeline with 10 user personas. We conduct a systematic evaluation of representative approaches on AirNav, ranging from traditional models to multimodal large language models (MLLMs), under unified metrics with open-source implementations. We further propose \textbf{AirVLN-R1}, trained via supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT), achieving state-of-the-art performance with a 51.82\% success rate on the test-unseen split. Real-world experiments on a physical UAV platform provide preliminary evidence of sim-to-real transferability, and our dataset and code are publicly available.