Dynamic Obstacle Avoidance in Blind Spots for Autonomous Mobile Robots Using CCTV Layer and Velocity Obstacle Method

This study proposes a method for dynamic obstacle detection and avoidance in blind spots beyond the sensing range of Autonomous Mobile Robots (AMRs). A top-down vision sensor (CCTV) monitors the driving environment where YOLO-based object detection identifies dynamic obstacles. The detected information is integrated into a custom CCTV layer and incorporated into the ROS2 navigation stack. Obstacle avoidance is then performed by the improvement DWA (Dynamic Window Approach) planner, enhanced with a Velocity Obstacle (VO) method utilizing the CCTV layer data. This approach extends the AMR’s field of view, both general and blind spot dynamic obstacles while generating more efficient global paths. The effectiveness of the proposed method is validated through experimental evaluation.