Integrity with LiDAR Incorrect Extraction Faults in Adverse Weather Conditions

This paper examines adverse weather impacts on the safety of LiDAR-based navigation for self-driving cars. Our prior work leveraged LiDAR intensity metrics to quantify and reduce incorrect feature extraction risk between pre-defined landmarks equipped with strong retro-reflectors and obstacles under a daytime scenario. However, LiDAR intensity is affected by weather conditions, showing correlations to water droplet density and solar radiation. We quantify the weather-affected LiDAR feature extraction risk using experimental data. Our results indicate that there is no significant increase in incorrect extraction risk under adverse weather conditions when the system incorporates landmarks with strong retro-reflectors that enhance the signal intensity returned to the LiDAR.