3-D mapping of an underground mall using a tracked vehicle with four sub-tracks

The authors attempted to create a 3-D map of an underground mall and subway station using a tracked vehicle. This paper is a field report of the 3-D mapping of the Sendai subway station by the tracked vehicle in Dec. 2007. From the ticket barriers to the platform, the Sendai subway station consists of three floors. For the 3-D mapping, we developed a tracked vehicle named ¿Kenaf¿, a small and light-weight 3-D laser scanner called TK scanner, and a robust 3-D scan matching method. Kenaf can pass through ticket barriers and climb up and down stairs, while TK scanner has a wide view angle and can measure dense 3-D shapes. During the experiment, the robot stopped at different points and collected 3-D scan data. The 3-D shapes were recorded by the TK scanner as point clouds. These 3-D point clouds were integrated into a map on the basis of odometry data on-line. The constructed map was not correct because of the lack of robot position z and the odometry error. The 3-D map was constructed by matching these 3-D point clouds off-line. To increase the robustness of the matching, we used the iterative closest point (ICP) matching method with a gravity constraint.