# AutoFly: Vision-Language-Action Model for UAV Autonomous Navigation in the Wild

arXiv: https://arxiv.org/abs/2602.09657 · HTML: https://mlsanity.com/paper/2602.09657/

## Key facts

- Cited by 15 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 10 Feb 2026.

## Built on by (most cited first, 15 of 15)

- [UAV-Track VLA: Embodied Aerial Tracking via Vision-Language-Action Models](https://mlsanity.com/paper/2604.02241/) (2026)
- [AsyncShield: A Plug-and-Play Edge Adapter for Asynchronous Cloud-based VLA Navigation](https://mlsanity.com/paper/2604.24086/) (2026)
- [FreeFly-Thinking : Aligning Chain-of-Thought Reasoning with Continuous UAV Navigation](https://mlsanity.com/paper/2603.07181/) (2026)
- [SpatialFly: Implicit 3D Prior-Guided Visual Reparameterization for Continuous UAV Vision-and-Language Navigation](https://mlsanity.com/paper/2603.21046/) (2026)
- [CosFly-VLA: A Spatially Aware Vision-Language-Action Model for UAV Tracking](https://mlsanity.com/paper/2607.15004/) (2026)
- [WaterVideoQA: ASV-Centric Perception and Rule-Compliant Reasoning via Multi-Modal Agents](https://mlsanity.com/paper/2602.22923/) (2026)
- [DynFly: Dynamic-Aware Continuous Trajectory Generation for UAV Vision-Language Navigation in Urban Environments](https://mlsanity.com/paper/2606.31654/) (2026)
- [Exp2VLA: Enabling Vision-Language-Action for Drone Navigation from Expert Demonstrations](https://mlsanity.com/paper/2607.03146/) (2026)
- [When Efficiency Becomes Fragility: Exploiting Dynamic Routing Vulnerabilities in Adaptive UAV Tracking](https://mlsanity.com/paper/2608.03902/) (2026)
- [LAPF: LLM-Agent-Based Path Finder Using the UAVScenes Dataset](https://mlsanity.com/paper/2608.15175/) (2026)
- [UAVs Meet Embodied Intelligence: Bridging Human Intents and Flying Dynamics Via Harnessing Physical-Digital AI Agents](https://mlsanity.com/paper/2609.18326/) (2026)
- [VLN on the Fly: An Onboard Vision-Language Navigation Stack for Aerial Robots](https://mlsanity.com/paper/2609.20191/) (2026)
- [DroneWAM: Efficient World Action Model for Drone Visual Navigation](https://mlsanity.com/paper/2609.33148/) (2026)
- [ForeFly: A Dual-Horizon World Action Model for Aerial Vision-Language Navigation](https://mlsanity.com/paper/2609.33581/) (2026)
- [From Local Whole-Body VLA Behaviors to Scene-Scale Aerial Manipulation](https://mlsanity.com/paper/2609.39670/) (2026)

## Abstract (arXiv)

Vision-language navigation (VLN) requires intelligent agents to navigate environments by interpreting linguistic instructions alongside visual observations, serving as a cornerstone task in Embodied AI. Current VLN research for unmanned aerial vehicles (UAVs) relies on detailed, pre-specified instructions to guide the UAV along predetermined routes. However, real-world outdoor exploration typically occurs in unknown environments where detailed navigation instructions are unavailable. Instead, only coarse-grained positional or directional guidance can be provided, requiring UAVs to autonomously navigate through continuous planning and obstacle avoidance. To bridge this gap, we propose AutoFly, an end-to-end Vision-Language-Action (VLA) model for autonomous UAV navigation. AutoFly incorporates a pseudo-depth encoder that derives depth-aware features from RGB inputs to enhance spatial reasoning, coupled with a progressive two-stage training strategy that effectively aligns visual, depth, and linguistic representations with action policies. Moreover, existing VLN datasets have fundamental limitations for real-world autonomous navigation, stemming from their heavy reliance on explicit instruction-following over autonomous decision-making and insufficient real-world data. To address these issues, we construct a novel autonomous navigation dataset that shifts the paradigm from instruction-following to autonomous behavior modeling through: (1) trajectory collection emphasizing continuous obstacle avoidance, autonomous planning, and recognition workflows; (2) comprehensive real-world data integration. Experimental results demonstrate that AutoFly achieves a 3.9% higher success rate compared to state-of-the-art VLA baselines, with consistent performance across simulated and real environments.