M.Sc. Nikita Bhardwaj Haupt
Wissenschaftliche Mitarbeiterin der Arbeitsgruppe Software Engineering: Dependability
M.Sc. Nikita Bhardwaj Haupt
Gebäude 32, Raum 429
Tel: +49 (631) 205-4471
Fax: +49 (631) 205-3331
Automated vehicles (AVs) have emerged as a promising means to reduce the high number of accidents on roads and improve the overall safety and driving experience for passengers. The development need for AVs has predominantly been driven by the high rate of accidents attributable to human error, which is responsible for more than 90\% of all road accidents. Nevertheless, a considerable challenge arises in how to ensure the safety of these vehicles. The transition from traditionally manually-driven vehicles to AVs presents substantial challenges for the prevailing safety assurance approaches within the automotive industry.
The automotive safety standard, ISO 26262, conducts a hazard analysis and risk assessment (HARA) process that relies on worst-case assumptions about the operational situation, resulting in unusually high safety integrity values. In many cases, this results in safety measures that require deactivating the currently activated functionality and returning control to the human driver. Moreover, focusing only on the most extreme situations results in overlooking the detailed complexities and dynamic characteristics of real-world environments. As a consequence, we get a robust system design with safety margins, guaranteeing that the system can endure extreme conditions while assuring safety. However, this comes at the expense of reduced performance and availability, and a static representation of the environment. Such an overly conservative approach towards safety assurance that assumes the human driver being the primary element of the vehicle control loop is not sufficient for AVs. As the driving automation increases, the involvement of the human driver in driving-related tasks decreases. In AVs, it is envisioned that the automated driving system (system) performs all dynamic driving tasks, maintains situational awareness, and allows the human driver to engage in non-driving-related tasks. Consequently, the responsibility for driving and safety shifts from the human driver to the system itself. Thus, to ensure that the system does not pose any unreasonable risk, it becomes crucial that it is self-aware, i.e., of its own functional capabilities and the current state of the situation in which it is operating. This can be accomplished by shifting certain aspects of safety assurance like risk assessment to runtime and allowing the vehicle to make safety-related decisions accordingly.
To this end, my research involves building a comprehensive framework Safety-rules-based runTime risk Assessment for aUTomated vehicleS (STATUS) that allows shifting the design time risk assessment to preventive risk assessment using a set of safety rules at runtime. With the help of STATUS, the framework incorporate the dynamic nature of the environment into hazard analysis and risk assessment. This approach shifts risk assessment from design time to runtime, permitting the vehicle to be self-aware of its associated risk. The components of the framework are spread across both design time and runtime. Design time activities include: (1) elhanced HARA (elHARA), (2) formal representation of knowledge about the system and operational situation, and (3) generating a set of safety rules. During runtime, system and situational parameters are monitored and the generated safety rules along with the represented knowledge are utilized to do the risk inference.