Learning-Based Autonomous UAV System for Electrical and Mechanical (E&M) Device Inspection

Area of Science:

• Robotics and Automation
• Computer Vision
• Aerospace Engineering

Background:

• Unmanned aerial vehicles (UAVs) offer flexibility and mobility for inspecting electrical and mechanical (E&M) devices.
• Traditional inspection methods can be inefficient and pose risks to human personnel.
• There is a growing need for autonomous systems to enhance inspection efficiency and safety.

Purpose of the Study:

• To present a UAV system capable of autonomously performing E&M device inspections.
• To develop a system that does not require prior knowledge of target device location or category.
• To validate the system's performance through real-world flight experiments.

Main Methods:

• Utilizing a learning-based detection approach for perception, leveraging semantic and spatial information from a 2-D object detector.
• Implementing multi-sensor fusion, combining visual data with depth measurements for accurate object state estimation.
• Employing advanced path planning algorithms to ensure fully autonomous inspection missions.

Main Results:

• The proposed UAV system successfully performed autonomous E&M device inspection missions.
• Flight experiments demonstrated stable and collision-free operation of the quadrotor platform.
• The system effectively integrated perception, localization, and path planning for autonomous operation.

Conclusions:

• The developed UAV system provides a viable solution for autonomous E&M device inspection.
• The system enhances inspection efficiency and reduces risks associated with manual inspections.
• This research contributes to the advancement of autonomous robotics in industrial applications.