A Dynamic Obstacle Avoidance Method in Robotic Navigation Control Through YOLOv5-Based Object Detection

The real-time obstacle avoidance strategy remains an important technical demand in robotic navigation. Although some progress has been achieved in recent years, existing works cannot handle unknown background inference in complex contexts. To deal with this challenge, this paper proposes a dynamic obstacle avoidance method in robotic navigation control through YOLOv5-based object detection. Specifically, an attention mechanism-based YOLOv5 algorithm is introduced to enhance the expression capability of small target channel features and spatial features. A P2 small target detection layer is added to strengthen the fusion of feature information between high and low layers. This is expected to address the problem that some detailed features of small targets are lost due to multiple down-sampling operations. For the navigation control part, an abstract potential field is introduced to guide the robots to move toward the target position while avoiding obstacles. Finally, we conduct some experiments on a real-world simulation scenario to make a comparative analysis. The obtained results show that the proposal can improve the obstacle avoidance effect compared with some baseline methods.