Warehouse Robot Detection for Human Safety Using YOLOv8

Abstract

Warehouse robots are commonly used by many companies such as Amazon. Depending on the technology within those robots, the safety of those that work around them can be in question. This paper presents a real-time object detection model that can detect and accurately follow the concurrent positions of i-Herb warehouse robots. The objective was to design an accurate YOLOv8-based model that could track each robot detected and be integrated with any camera system to protect the humans that work around the robots. The experimental results show that the model can detect robots with an accuracy of 98.9%. This is an extremely high accuracy with a slight asterisk next to it due to the limited amount of data for this experiment, even with the use of data augmentation within Roboflow. As a proof of concept, this work proves that the robots have features that are easily recognized and trained for real-time object detection, and with further data, the model would be accurate and robust in any environment.