Automated Accessibility Map Construction with Two-Level Viewpoint Planning Strategy

Accessibility maps are widely used for providing access information to users of wheelchairs and strollers. In an accessibility map, the location of ramps and elevators for each building is essential for planning a route, for example, from the entrance gate to a destination building. SLAM (simultaneous localization and mapping) is a standard tool for automatically making a geometrical map by a mobile robot. Once we have a map, we can manually add ramp information afterward, but this operation is inefficient, and the resultant map may not be accurate. Therefore, we develop a method of generating a SLAM map with ramp information together. Ramp structures are usually smaller than building structures, and observing the target environment (e.g., a university campus) with a single resolution is inappropriate. We thus propose a two-level viewpoint planning approach where the robot observes only ramp regions in a denser resolution. With the test in a large simulated environment, the proposed method shows a better performance with a small extra cost than the method without the viewpoint planning for ramp detection.