# CosFly-VLA: A Spatially Aware Vision-Language-Action Model for UAV Tracking

arXiv: https://arxiv.org/abs/2607.15004 · HTML: https://mlsanity.com/paper/2607.15004/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Aerial and drone VLA, one of 13 directions in Vision-language-action models.
- Builds on 48 of its 67 references that are in the map.
- First posted on arXiv 16 Jul 2026.

## Placed in

- Vision-language-action models / [Aerial and drone VLA](https://mlsanity.com/vla/aerial-and-drone-vla/)

## Builds on (most cited first, 30 of 48)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [RT-1: Robotics Transformer for Real-World Control at Scale](https://mlsanity.com/paper/2212.06817/) (2022)
- [Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware](https://mlsanity.com/paper/2304.13705/) (2023)
- [Alpamayo-R1: Bridging Reasoning and Action Prediction for Generalizable Autonomous Driving in the Long Tail](https://mlsanity.com/paper/2511.00088/) (2025)
- [A Survey on Vision-Language-Action Models for Autonomous Driving](https://mlsanity.com/paper/2506.24044/) (2025)
- [AutoTrack: Towards High-Performance Visual Tracking for UAV with Automatic Spatio-Temporal Regularization](https://mlsanity.com/paper/2003.12949/) (2020)
- [UAVs Meet LLMs: Overviews and Perspectives Toward Agentic Low-Altitude Mobility](https://mlsanity.com/paper/2501.02341/) (2025)
- [TrackVLA: Embodied Visual Tracking in the Wild](https://mlsanity.com/paper/2505.23189/) (2025)
- [A Dual Process VLA: Efficient Robotic Manipulation Leveraging VLM](https://mlsanity.com/paper/2410.15549/) (2024)
- [Open3D-VQA: A Benchmark for Comprehensive Spatial Reasoning with Multimodal Large Language Model in Open Space](https://mlsanity.com/paper/2503.11094/) (2025)
- [End-to-end Active Object Tracking and Its Real-world Deployment via Reinforcement Learning](https://mlsanity.com/paper/1808.03405/) (2018)
- [UAV-Flow Colosseo: A Real-World Benchmark for Flying-on-a-Word UAV Imitation Learning](https://mlsanity.com/paper/2505.15725/) (2025)
- [UAV-VLA: Vision-Language-Action System for Large Scale Aerial Mission Generation](https://mlsanity.com/paper/2501.05014/) (2025)
- [UnrealZoo: Enriching Photo-realistic Virtual Worlds for Embodied AI](https://mlsanity.com/paper/2412.20977/) (2024)
- [Similarity-Guided Layer-Adaptive Vision Transformer for UAV Tracking](https://mlsanity.com/paper/2503.06625/) (2025)
- [VLA-AN: An Efficient and Onboard Vision-Language-Action Framework for Aerial Navigation in Complex Environments](https://mlsanity.com/paper/2512.15258/) (2025)
- [CognitiveDrone: A VLA Model and Evaluation Benchmark for Real-Time Cognitive Task Solving and Reasoning in UAVs](https://mlsanity.com/paper/2503.01378/) (2025)
- [HRVQA: A Visual Question Answering Benchmark for High-Resolution Aerial Images](https://mlsanity.com/paper/2301.09460/) (2023)
- [AirCopBench: A Benchmark for Multi-drone Collaborative Embodied Perception and Reasoning](https://mlsanity.com/paper/2511.11025/) (2025)
- [RefDrone: A Challenging Benchmark for Referring Expression Comprehension in Drone Scenes](https://mlsanity.com/paper/2502.00392/) (2025)
- [RSPT: Reconstruct Surroundings and Predict Trajectories for Generalizable Active Object Tracking](https://mlsanity.com/paper/2304.03623/) (2023)
- [AutoFly: Vision-Language-Action Model for UAV Autonomous Navigation in the Wild](https://mlsanity.com/paper/2602.09657/) (2026)
- [WorldVLN: Autoregressive World Action Model for Aerial Vision-Language Navigation](https://mlsanity.com/paper/2605.15964/) (2026)
- [Learning Vision-based Pursuit-Evasion Robot Policies](https://mlsanity.com/paper/2308.16185/) (2023)

## Built on by (most cited first, 1 of 1)

- [UAVs Meet Embodied Intelligence: Bridging Human Intents and Flying Dynamics Via Harnessing Physical-Digital AI Agents](https://mlsanity.com/paper/2609.18326/) (2026)

## Abstract (arXiv)

Dynamic target tracking is essential for Unmanned Aerial Vehicles (UAVs) operating in complex urban environments, where both the target and the camera viewpoint change continuously. Existing Vision-Language-Action (VLA) policies can track visible targets effectively, but their performance often degrades when buildings, vegetation, or roadside objects block the line of sight. During sustained occlusion, a policy may lose the target state, execute actions toward an incorrect region, and amplify this error through subsequent observations until re-acquisition becomes impossible. To this end, we present CosFly-VLA, a spatially aware VLA model that jointly grounds the target, estimates its visibility, and generates continuous flight actions through a structured prediction interface. To train this policy, we use a large-scale recipe over diverse data sources. Spatially Grounded Continued Pretraining (CPT) on a 500k mixed pool injects UAV-view depth, distance, and 3-D spatial reasoning. A three-stage Curriculum-based Supervised Fine-Tuning (SFT) process then specializes the tracker through multi-head warm-up followed by two-stage curriculum learning over natural and hard / long-occlusion data. Chain-of-Thought (CoT) training subsequently teaches recovery-oriented reasoning traces before structured answers. Finally, a closed-loop Reinforcement Learning (RL) stage optimizes tracking behavior with a multi-component reward covering stand-off tracking, grounding quality, collision avoidance, and task success. Relative to OpenVLA, CosFly-VLA-0.8B reduces open-loop Average Displacement Error (ADE) by 34.1% on seen-test and 35.3% on unseen-test. Closed-loop optimization improves Success Rate (SR) by 29.8% and 2.5%, respectively. These results demonstrate progress from visible-frame imitation toward spatially grounded action-closed-loop control, evaluated under a shared oracle state history.