Autonomous Multi-UAV System for Runway FOD Detection Using Hybrid Navigation

This paper presents a novel multi-agent UAV-based framework for detecting and localizing foreign objects on airport runways, addressing critical aviation safety and operational efficiency concerns. Traditional methods such as manual inspection, vehicle-based patrols, radar systems, infrared sensors, and fixed cameras suffer from limitations that include slow operation or high costs or elevated false positive rates for small object detection. To overcome these challenges, we propose an integrated approach leveraging deep learning for both object detection and visual servoing. We have developed a custom dataset specifically designed for runway foreign object detection, localization, and visual servoing tasks. Our system implements a hybrid navigation strategy combining visual feedback and GPS, utilizing limited but reliable runway features such as threshold and edge markings. The framework employs a multi-agent approach for runway inspection, which has been successfully validated using two UAVs with demonstrated potential for system scalability. This methodology significantly reduces inspection time while maintaining the capability for human intervention when required. We present comprehensive experimental validation conducted under outdoor conditions, accompanied by detailed results and analysis that demonstrate the effectiveness of our approach.