Technical Aspects of Deploying UAV and Ground Robots for Intelligent Logistics Using YOLO on Embedded Systems

Area of Science:

• Computer Vision
• Artificial Intelligence
• Robotics

Background:

• Logistics automation leverages image processing and AI for efficiency and resilience.
• Deploying deep learning on resource-constrained embedded platforms for real-time logistics is challenging.

Purpose of the Study:

• To compare YOLOv5 and YOLOv8 performance for object detection on embedded platforms in logistics.
• To develop a user-friendly GUI for real-time detection result visualization and filtering.
• To provide deployment guidelines for UAV- and ground robot-based logistics.

Main Methods:

• Comparative analysis of YOLOv5 and YOLOv8 using COCO and a custom logistics dataset.
• Evaluation of inference speed, accuracy, and dataset-specific metrics.
• Development of a graphical user interface (GUI) for object visualization.
• Investigation of deployment strategies in Python and C# environments.

Main Results:

• Performance metrics (speed, accuracy) for YOLOv5 and YOLOv8 on embedded systems.
• Demonstration of the GUI's effectiveness in real-time object visualization and filtering.
• Comparative analysis of Python and C# deployment impacts on performance and scalability.

Conclusions:

• YOLOv8 shows potential for enhanced logistics automation on embedded platforms.
• The developed GUI significantly improves the practical usability of object detection systems.
• Optimized deployment strategies are crucial for efficient resource utilization in robotic logistics.