AerialVLA: A Vision-Language-Action Model for UAV Navigation via Minimalist End-to-End Control

Researchers propose AerialVLA, a minimalist end-to-end Vision-Language-Action framework for UAV navigation that directly maps visual observations and linguistic instructions to continuous control signals. The system eliminates reliance on external object detectors and dense oracle guidance, achieving nearly three times the success rate of existing baselines in unseen environments.

• →AerialVLA introduces a streamlined dual-view perception strategy that reduces visual redundancy while preserving essential navigation cues.
• →The framework deploys fuzzy directional prompting using only onboard sensors, eliminating dependency on external oracle guidance.
• →The system integrates continuous 3-DoF kinematic commands with intrinsic landing signals for autonomous precision landing.
• →Testing on TravelUAV benchmark shows state-of-the-art performance in seen environments and superior generalization in unseen scenarios.
• →The minimalist approach demonstrates that end-to-end systems can capture more robust visual-motor representations than complex modular systems.