Fail Safe Process of Vehicle Localization for Reliability Improvement of LV3 Autonomous Driving

Abstract

In this paper, a vehicle localization fail safe process is proposed for improving localization reliability for level 3 autonomous driving. This process does not necessitate additional and expensive sensor configuration using sensor fusion with practical usage and high density maps for localization fail safe. The proposed process also suggests a three-step safety mechanism. The first step is to detect and monitor in-vehicle sensors. The second step constitutes Dead Reckoning (DR) model-based fail monitoring. The final is a map-matching fail safe to detect and to identify fail level and this algorithm recovers abnormal positions of map-matching results. The fail detection algorithm and monitoring logic and thresholds were validated and identified by vehicle endurance run tests comprising over 100,000 km of driving. The performance of DR fail monitoring and map-matching base fail logic were evaluated by vehicle-simulations on sensor measurements. The results demonstrated that the proposed process achieves improvement of reliability with accuracy fault detection and identification for abnormal cases on the fail level.