Sensor Data Fusion of Intelligent Autonomous Mover for Object Detection and Collision Avoidance in Environments with Surrounding Crowds

摘要

In this paper, a sensor data scheme, which combines various types of sensing data for an autonomous mover to achieve object detection & classification, collision avoidance, is presented. The sensing data include distance and DoA information from an ultrasound radar array, point cloud data from a 1-ray LiDAR, and AI-powered image based sensing results from a webcam. Useful environment messages such as the locations (distances and directions), the moving speeds, and the facial directions of the surrounding pedestrians can be extracted after fusion. Since image based sensing does not provide depth information, the fusion tries to find its mapping to the 2-D data from LiDAR and ultrasound radar array. The object distance is used as the major fusion index and data calibration across different sensing devices are performed. A danger model is associated with each detected pedestrian based on the fusion results. This facilitates social aware navigations in indoor people rich environments such as supermarkets. Experimental results are presented in this paper and comparisons with previous work for similar autonomous mover applications are also provided.