Basic Experiment of LIDAR Sensor Measurement Directional Instability for Moving and Vibrating Object

In this paper, a basic study on directional instability of measured data in a vehicle equipped with a LIDAR sensor. To realize autonomous car control without human, it is important to estimate the correct current position and it would be measured by LIDAR device generally. The difficulty in the process is the angle estimation. In rotational motion, all points in the point cloud data shift to different distances and directions, making the estimation unstable.We improved ICP to remove the instability of angle estimation. One is the two-sided search method, which improves the way points correspond in ICP. The other one is to convert the shape data (point cloud data) to the r-6 coordinate system so that the point cloud can easily obtain the best correspondence. We conducted estimation using point cloud data that imitate a road with fences placed on both sides. As a result, the range of accuracy within 0.1 degrees of the estimation error for the twosided search method was 0.63 times larger than that for ICP. Similarly, in the case of ICP in the r-6 coordinate system, the result was 1.76 times larger. Similarly, in the case of ICP in the r6 coordinate system, the result was 1.76 times larger. An analysis of the evaluation function E revealed the reason why the twosided search method did not estimate well. It was because each point prevented the selection of the correct correspondence.These results indicate that ease of response is important in self-location estimation. However, the r-6 coordinate system is not suitable for estimating movement, we would like to improve it so that it can be used in all situations.