SIMULTANEOUS LOCALIZATION AND MAPPING FOR SEMI-SPARSE POINT CLOUDS

Abstract. 3D representation of the environment is a piece of vital information for most of the engineering sciences. However, providing such information in classical surveying approaches demands a considerable amount of time for localizing the sensor in a desired coordinate frame to map the environment. Simultaneous Localization And Mapping (SLAM) algorithm is capable of localizing the sensor and do the mapping while the sensor is moving through the environment. In this paper, SLAM will be applied on the data of a lightweight 3D laser scanner in which we call semi-sparse point cloud, because of the unique specifications of the point cloud which comes from various resolutions in vertical and horizontal directions. In contrast to most of the SLAM algorithms, there is no aiding sensor to provide prior information of motion. The output of the algorithm would be a high-density full geometry detailed map in a short time. The accuracy of the algorithm has been estimated in a medium scale simulated outdoor environments in Gazebo and Robot Operating System (ROS). Considering Velodyne Puck accuracy which is 3 cm, the map was generated with approximately 6 cm accuracy.