Ontology-Enhanced STPA Method of Scenario Safety Constraint Identification for Autonomous Driving

Hazards from complex operational scenarios bring huge challenges for autonomous driving. This study proposes the External Operational Scenario-Systems Theoretic Process Analysis (EOS-STPA) approach, which extends the scope of traditional Systems Theoretic Process Analysis (STPA) to operational scenarios. By incorporating ontology-based knowledge representation and employing a hierarchical control structure that encompasses scenario control actions with feedback loops to ensure adaptive decision-making, EOS-STPA can formalize the interaction between systems and operational scenarios. Furthermore, EOS-STPA identifies and generates formalized scenario safety constraints comprehensively while enhancing safety of the intended functionality (SOTIF) for AD. Additionally, EOS-STPA’s hierarchical control modeling facilitates efficient scenario hazard identification through structured scenario decomposition. In the context of an autonomous vehicle car-following scenario, EOS-STPA demonstrates its strong capability in formalized safety analysis. This study is a pioneering effort to extend STPA to external operational scenarios while transforming technical system perspectives into operational scenario viewpoints.