Can 3D Lidar localize when placed vertically ?
Currently the Velodyne puck being used on our UAV can localize in an obstacle course room when it is mounted with the base parallel to the ground. But there are more obstacles above and below the robot, so the UAV will be able to see them better if we mounted it such that the base was at 90 degrees to the ground. But when this is done, there is no localization using ethz-asl ICP mapper. As a result no maps are built.
I guess that it is because when mounted vertically the robot is not able to see many of the (older) features it already passed by in the current frame. Is it not possible for scan matching to identify the current position based on the last two frames? Because there is definitely an overlap of features between the current frame and the last 3-5 frames.
