Integrity with Extraction Faults in LiDAR-Based Urban Navigation for Driverless Vehicles.

This paper examines the safety of LiDAR-based navigation for driverless vehicles and aims to reduce the risk of extracting information from undesired obstacles. We define the faults of a LiDAR navigation system, derive the integrity risk equation, and suggest landmark environments to reduce the risk of fault-free position error and data association faults. We also present a method to quantify feature extraction risk using reflective tape on desired landmarks to enhance the intensity of returned signals. The high-intensity returns are used in feature extraction decisions between obstacles and pre-defined landmarks using the Neyman-Pearson Lemma. Our experiments demonstrate that the probability of incorrect extraction is below 10(-14), and the method is sufficient to ensure safety.