Abstract
In this paper, we propose techniques which make estimation of traversability accurately with stereo camera for autonomous mobile robot navigation. In order to achieve the purpose, we employ the approach which integrates three kinds of filters. These filters can remove disparity errors which result from incorrect correspondence between stereo images. As a result, only data with high reliability remains. And we develop a stereo vision system to measure accurate geometrical information of terrain. The data of 3-dimensions acquired by this approach is changed into the grid map of 2-dimensions. This grid map is used to estimate traversability, and it can also be used for path planning. We present and evaluate experimental results for our system in outdoor environment. Experimental results show that our technique can remove disparity errors correctly. And filtered grid map can estimate Traversability accurately than unfiltered grid map.
