OntoSce: Domain-Oriented FTA Analysis for Generating Safety-Critical Scenarios

The widespread application of autonomous driving technology has triggered in-depth discussions on its safety and reliability. To address this challenge, this paper presents the safety-critical scenario generation tool OntoSce, which is driven by a domain model and integrates Fault Tree Analysis (FTA) for causal analysis, enabling the automated generation of safety-critical scenarios. The research first explores domain-oriented FTA analysis, elaborating on two dimensions: traffic rules and hazardous scenarios. Specifically, FTA analysis is performed using the knowledge from the constructed domain model to derive key objects and their environmental conditions that may lead to accidents or violations. By integrating multiple influencing factors, OntoSce can generate diverse scenarios, providing an important data foundation for the safety assessment of autonomous driving technology. We conducted experiments using the scenarios generated by OntoSce in the Apollo simulation environment, successfully identifying safety violations. This validated the feasibility and effectiveness of the generated scenarios. Additionally, the scenarios generated by OntoSce are compatible with all simulation platforms, such as CARLA, LGSVL, etc.