This paper details a system for an autonomous navigation of electric wheelchair in outdoor urban area. The goal for this study is to achieve autonomous navigation on paved roads including the slopes in a busy street. We propose the autonomous navigation system using '2D drivable map' that explains steps, obstacles and paved/unpaved areas. 2D drivable map generated from 3D point cloud map by robot itself. We detected smoothly connected planes and obstacles by region growing segmentation based on normal in 3D point cloud map and projected the vertically onto a 2D map. A paved/unpaved area segmentation is based on intensity of 2D map. The experimental results show the 2D drivable map correctly explains steps and obstacles in urban areas and detects changes from roads to lawns. An electric wheelchair achieved autonomous navigation in urban areas without colliding with obstacles or mounting lawns.