Development of an Autonomous Unmanned Aerial Manipulator Based on a Real-Time Oriented-Object Detection Method

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Autonomous Unmanned Aerial Manipulators (UAMs) show potential for 3D grasping but face challenges in object detection and indoor positioning.
• Efficient and precise autonomous grasping requires robust object recognition and pose estimation.

Purpose of the Study:

• To develop a novel oriented-object detection method for UAMs.
• To design and implement an integrated UAM platform for autonomous grasping in GPS-denied environments.

Main Methods:

• Introduced Rotation-SqueezeDet, an oriented-object detection method for embedded platforms, providing rotation-aware bounding boxes.
• Developed a UAM system with detailed formulation, positioning, control, and planning.
• Open-sourced all mechanical designs for community reuse.

Main Results:

• Rotation-SqueezeDet achieves near real-time performance on embedded systems.
• The integrated UAM system demonstrated effective autonomous aerial rotational grasping.
• Experimental trials validated the system's applicability in GPS-denied environments.

Conclusions:

• The proposed Rotation-SqueezeDet method and UAM platform significantly advance autonomous grasping capabilities.
• The open-source hardware design promotes wider adoption and development in the robotics community.
• The system is suitable for various industrial applications requiring complex manipulation tasks in challenging environments.