Enhanced Laser-Scan Matching with Online Error Estimation for Highway and Tunnel Driving

Lidar data can be used to generate point clouds for the navigation of autonomous vehicles or mobile robotics platforms. Scan matching, process estimating rigid transformation that best aligns two clouds, is basis lidar odometry, a form dead reckoning. odometry particularly useful when absolute sensors, like GPS, are not available. Here we propose Iterative Closest Ellipsoidal Transform (ICET), scan matching algorithm which provides novel improvements over current state-of-the-art Normal Distributions (NDT). Like NDT, ICET decomposes into voxels and fits Gaussian distribution points within each voxel. The first innovation reduces geometric ambiguity along large flat surfaces by suppressing solution those directions. second infer output error covariance associated with position orientation between successive clouds; incorporated state-estimation routine such as an extended Kalman filter. We constructed simulation compare performance NDT in 2D space both without found produces superior estimates while accurately predicting accuracy.