Enhancing Collision Avoidance in Mobile Robots Using YOLOv5: A Lightweight Approach for Unstructured Environments

Mobile robots play a crucial role in Industry 4.0, particularly in dynamic and unstructured environments where moving obstacles present significant challenges. This study applies the YOLOv5 object detection algorithm to enhance robotic perception and obstacle avoidance. The primary objective is to improve the accuracy and speed of object detection in real-time scenarios, ensuring safer and more efficient navigation for robots. The research contribution lies in developing a lightweight YOLOv5 model optimised for robotic applications, capable of detecting objects with high accuracy. The model was trained on a diverse dataset of 10,700 images, including static and dynamic objects such as chairs, fans, fire extinguishers, and humans, captured under various conditions and orientations. The dataset was divided into training (70%), validation (15%), and testing (15%) subsets. The proposed model achieved a mean average precision (mAP) of 0.73 at a confidence threshold of 0.374, demonstrating superior performance compared to the YOLOv4 model in terms of accuracy and processing speed. Notably, the model excelled in detecting static objects such as chairs, achieving a perfect recognition rate of 1.00, while encountering challenges with moving objects such as humans due to motion blur and rapid changes in body posture. These findings highlight the model’s potential for real-time applications in industrial and unstructured environments. In conclusion, this study demonstrates that the enhanced YOLOv5 model significantly improves object detection and collision avoidance capabilities in robotic systems.