Robust 6DoF Localization in Dynamic Urban Environment ?

摘要

In this paper, we propose self-position estimation method considering dynamic obstacles. We obtain the probability that a voxel is containing a static obstacle, based on the number of times the voxel has been occupied by the obstacle. Then we give this probability to the point cloud in the voxel and perform robust and highly accurate relative posture estimation with the normal of this point. Finally, we propose a method to stably perform self-position estimation of 6 degrees of freedom by performing translational alignment using ICP.