Recently, technologies related to autonomous vehicles are researched all over the world. For such autonomous vehicles, it is important to extract small obstacles such as curb stone, and it is also necessary to recognize drivable area in omni-direction not affected by occlusions. In this report, we propose obstacle extraction method based on Occupancy Grid Maps to overcome these problems. Moreover, the system overview of our autonomous vehicle will be shown and result of proposed algorithm at a demonstration event in the Tokyo motor show 2011 will also be shown.