Airborne Visual Detection and Tracking of Cooperative UAVs Exploiting Deep Learning.

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Artificial Intelligence

Background:

• Coordinated Unmanned Aerial Vehicle (UAV) formations, or swarms, offer significant performance benefits for civil and military applications.
• Effective coordination relies on UAVs visually monitoring each other, necessitating robust methods for detecting and tracking cooperative targets in image sequences.
• Existing solutions face challenges with varying illumination, backgrounds, and target distances.

Purpose of the Study:

• To develop an innovative deep learning-based approach for detecting and tracking cooperative targets within UAV formations.
• To enhance the coordination capabilities of UAV swarms through improved machine vision algorithms.
• To validate the proposed method using real-world flight test data.

Main Methods:

• Integration of the You Only Look Once (YOLO) object detection system into a novel processing architecture.
• Leveraging navigation hints from the cooperative nature of the UAV formation to aid machine vision algorithms.
• Conducting an experimental flight test campaign with multirotor UAVs to gather image data.

Main Results:

• The proposed approach demonstrated high-level accuracy in detecting and tracking UAVs within formations.
• The system exhibited robustness against challenging environmental conditions, including variable illumination, complex backgrounds, and changing target ranges.
• Experimental validation confirmed the effectiveness of the integrated YOLO system and navigation hint strategy.

Conclusions:

• The developed deep learning framework significantly improves the ability of UAVs to monitor formation members.
• This technology enhances the robustness and performance of coordinated UAV operations in diverse scenarios.
• The findings pave the way for more sophisticated autonomous coordination in multi-UAV systems.