RecoverFly: A Failure-Aware Reinforcement Learning Post-Training Framework for Aerial Vision-Language Navigation
2026-08-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionArtificial Intelligence
AI summaryⓘ
The authors address how drones can better follow language instructions while flying in complex environments. They note that existing drone policies trained by copying behavior struggle with adapting during real flights. To improve this, the authors design RecoverFly, a reinforcement learning method that focuses on fixing past mistakes, handling difficult scenes, and keeping learned skills intact. Their experiments show that RecoverFly helps drones succeed more often and generalize better to new places and objects using fewer training steps. This work helps drones understand and act on language instructions more reliably.
Unmanned Aerial Vehicle (UAV)Vision-Language Navigation (VLN)Reinforcement Learning (RL)Behavior CloningPolicy OptimizationAutoregressive ModelsCurriculum LearningGeneralizationFlight ControlInteractive Robotics
Authors
Boxiong Wang, Hui Kang, Geng Sun, Jiahui Li, Chao Yu, Daxin Tian
Abstract
Unmanned aerial vehicle vision-language navigation (UAV-VLN) requires agents to translate visual observations and language instructions into reliable flight actions in complex environments. Although recent end-to-end UAV vision-language-action (UAV-VLA) policies reduce reliance on separately designed perception, planning, and control modules, their behavior-cloning objectives provide limited corrective supervision for interactive closed-loop execution. Reinforcement learning (RL) offers a promising solution, while its effectiveness is constrained by inefficient use of samples, long-tailed scene distributions, and policy distribution shift during optimization. To this end, we propose RecoverFly, a failure-aware RL post-training framework for end-to-end UAV-VLA policies. Specifically, RecoverFly adapts token-level RL for stable optimization of grammar-constrained autoregressive UAV actions, revisits unresolved failure cases to strengthen corrective learning and sample utilization, and combines a two-stage long-tail scene curriculum with reference-policy regularization to improve scene adaptation while preserving acquired capabilities. Experiments on the TravelUAV benchmark demonstrate that RecoverFly achieves the best performance on the seen, unseen-map, and unseen-object splits. Moreover, compared to the AerialVLA initialization, RecoverFly improves success rate by 3.12 to 8.37 percentage points under a total rollout budget of about 30\% of the training-set size, validating its effectiveness, robustness, and generalization capabilities.