A Comprehensive Failure Analysis Method for a Fusion Localization System: A Case Study on Vector High-Definition Map and Vision Matching

For automated driving technology with a fusion localization system, especially the matching of vector high-definition (HD) maps with vision data for localization, the complexity of the localization system makes it challenging to accurately identify the triggering conditions of failure. Therefore, a comprehensive failure analysis method is proposed that combines a seven-step qualitative analysis with hierarchical quantitative testing processes. First, through leveraging the forward analysis characteristics of Hazard and Operability (HAZOP) analysis and the reverse analysis characteristics of Systems-Theoretic Process Analysis (STPA), a bidirectional closed-loop failure analysis process is designed to ensure the completeness of the analysis. Then, real vehicle testing is hierarchically decoupled into the open-loop localization test and the closed-loop control test. In particular, feature parameters are designed to simulate triggering conditions and are injected into the localization system to test the accuracy of the qualitative results. Finally, a vector HD map and vision matching (VHMVM)-based localization system is adopted as a specific case to the verify failure analysis method. The results show that the proposed failure analysis method can ensure the completeness and accuracy of the analysis as much as possible and obtain trustworthy triggering conditions.