Short-range and Long-range Obstacle Detection Method for a Delivery Robot Based on Multi-sensor Fusion

The development of an effective obstacle perception system is critical for preventing potential collisions between an autonomous delivery vehicle and obstacles in its path. The position of obstacles, which can be determined by their 3D location and yaw value, is vital in facilitating reliable path planning for the vehicle. However, most conventional approaches to obstacle detection rely on a single sensory system, leading to blind spots where obstacles may go undetected due to hardware limitations. This paper proposes a novel approach that fuses three sensors - rotating LiDAR, horizontal LiDAR, and a camera sensor - to create a robust obstacle detection system. This new system enables the detection of short-and long-range obstacles previously undetectable due to hardware limitations. The camera sensor is also utilized to classify the detected objects, thereby enhancing the overall reliability of the perception system. The paper proposed a unique fusion method to detect and classify obstacles for the delivery vehicle.