Research on the SOTIF Analysis Method for Autonomous Perception Function Based on Adapted STPA

With the development of autonomous systems, the occurrence of accidents caused by perception biases has become a significant concern. To comprehensively identify the causal factors of SOTIF hazards induced by perception systems, this study proposes an adapted System-Theoretic Process Analysis (STPA) method that improves the structure of the sensor module in the control loop, which maps the hazard analysis process to the risk-triggering mechanism. The adapted STPA considers both perception devices and perception algorithms and incorporates guiding words from the UL4600 standard to thoroughly identify causal factors. A practical case study of an anti-pinch system is presented as an example of the proposed method, where the autonomous perception module was introduced into the system. The triggering events identified by the proposed approach indicate that the new method improves the knowledge about SOTIF-related hazards and tackles the safety constraints of the system. The adapted STPA offers better modeling and assessment of autonomous perception function, which could support the evaluation of autonomous systems’ architectural design and build the SOTIF concept.