Enhanced Integrity of Lidar Localization: A Study on Feature Extraction Techniques

This study develops a methodology to minimize extraction faults in landmark-based Lidar localization using a map. The analysis begins by addressing the integrity risks associated with Lidar navigation systems in self-driving cars and identifying the causes of incorrect extractions. Example Lidar-fault integrity requirements are introduced to establish a basis for data point selection using the Neyman-Pearson criterion. Weighted least squares estimation, combined with a chi-square residual test, is then applied for landmark extraction. Experimental results confirm that the methodology reduces landmark extraction errors to centimeter-level.