LiDAR-to-Map Registration: Comparative Analysis of Mechanical and Solid-State LiDAR Technologies Across ICP and NDT Algorithms

Accurate positioning is essential for autonomous vehicle (AV) navigation systems, supporting tasks such as motion planning, decision-making, and control. Although the Global Navigation Satellite System (GNSS) is widely used to provide positioning services, its accuracy and reliability can degrade and may even become unavailable in urban and indoor environments. However, AVs must have access to a positioning solution in all environments at all times. To bridge occurrences of GNSS unreliability, researchers have investigated the use of perception sensors such as LiDAR to sense the environment around the AV and provide an alternative positioning solution. LiDAR-based positioning methods, including LiDAR odometry (LO) and map matching, typically rely on registration algorithms such as Iterative Closest Point (ICP) and Normal Distribution Transform (NDT) for pose estimation. Mechanically Spinning LiDAR (MSL) is a well-established LiDAR technology that has been mounted on AVs to assist in tasks such as positioning and mapping. More recently, Solid-State LiDAR (SSL) has emerged as a promising new LiDAR technology, offering advantages over MSL in the form of fewer moving parts and lower costs. This study introduces a LiDAR-to-Map Registration (LMR) pipeline designed to evaluate the performance of MSL and SSL for AV positioning using real-world data. The comparative analysis investigates the suitability of each LiDAR technology accross various dynamic driving scenarios and through diverse environments, highlighting their strengths in the scenarios challenging to GNSS positioning.