The Construction of Occupancy Grid Map with Semantic Information for the Indoor Blind Guiding Robot

Most of the blind people need to be accompanied when walking indoors, which limits the activities of the blind and increases the cost of labor. We designed a low-cost indoor guide robot equipped with two-dimensional Lidar and depth camera to complete the construction of indoor occupancy grid map with semantic. The Cartographer algorithm was used to construct occupancy grid map and avoid noise problems caused by crowds during map construction. Obstruct detector tool was used to obstruct dynamic obstacles, which were removed from the lidar data to achieve good map construction. At the same time, we use MobileNet-SSD to detect the categories of common indoor obstacles. Combined with the ranging information of depth camera and the positioning information of SLAM, we can obtain the world coordinates of obstacles. Therefore, the obstacle information can be marked on the occupancy grid map constructed by SLAM, and a occupancy grid map with semantic information can be obtained. Through the interactive way of voice module, the blind person is informed of the category of obstacles and distance information, so as to ensure the safety of the blind person and good user experience.