Layered SOTIF Analysis and 3σ-Criterion-Based Adaptive EKF for Lidar-Based Multi-Sensor Fusion Localization System on Foggy Days

The detection range and accuracy of light ranging (LiDAR) systems are sensitive to variations in fog concentration, leading the safety intended functionality-related (SOTIF-related) problems LiDAR-based fusion localization system (LMSFLS). However, due uncontrollable weather, it is almost impossible quantitatively analyze effects on LMSFLS a realistic environment. Therefore, this study, we conduct layered quantitative SOTIF analysis foggy days using simulation. Based results, identify component-level, system-level, vehicle-level functional insufficiencies LMSFLS, corresponding triggering conditions, potential SOTIF-related risks. To address risks, propose modification strategy that incorporates visibility recognition 3σ-criterion-based variance mismatch degree grading adaptive extended Kalman filter. scenario recognized judge whether measurement information LiDAR odometry disturbed by fog. Moreover, proposed filter adopted fuse abnormal with IMU GNSS. Simulation results demonstrate can inhibit divergence improve self-driving cars days, accurately recognize scenarios.