Runtime verification safety engineering methodology and proof of concept for a connected autonomous shuttle at pedestrian crossings in a Bus Rapid Transit (BRT) system.
Last modified:
What is it about:
Summary of the project
The challenge:
What was the challenge that the company approached IESE with?
The support:
How did IESE specifically support the company?
The result:
What are the most important results of the project and what are the next steps for the company?
Your benefits:
Fraunhofer IESE is your partner for safety engineering
In Hitachi-City, Japan, autonomous shuttles are poised to enhance the Bus Rapid Transit (BRT) system. Hitachi-City aims to boost the efficiency and reliability of public transportation while also addressing the future scarcity of qualified drivers. These autonomous shuttles operate on dedicated BRT roads, ensuring a smooth and uninterrupted journey, isolated from regular traffic. However, the BRT roads do intersect with regular traffic lanes and pedestrian crossings, increasing the potential risk of collisions and thus harm to both traffic participants and bus passengers.
Recognizing these challenges, Hitachi, Ltd. has, in partnership with Fraunhofer IESE, developed a cutting-edge dynamic safety engineering process. This innovative solution enhances the safety of autonomous shuttles, particularly in interactions with pedestrians and other traffic participants. The system has been developed meticulously and systematically and has been successfully evaluated at the Hitachi campus in Japan, proving its effectiveness and scalability in the BRT system and other real-world applications.
Note: Hitachi, Ltd., and Hitachi-City BRT are two separate independent entities. The use case considered is an example scenario to show the applicability of our solution.
The primary scientific challenge lies in developing autonomous shuttles that not only enhance passenger benefits but also ensure safety. The Bus Rapid Transit (BRT) road features numerous unsignalized and some signalized pedestrian crossings, which present complex and dynamic scenarios for autonomous navigation due to frequent vehicle-pedestrian interactions. Traditional approaches often involve applying worst-case environmental assumptions during system operation, particularly concerning pedestrian behavior. This method results in overly conservative responses at pedestrian crossings, leading to performance that falls short of that of human drivers. Such cautious behavior can reduce the effectiveness, utility, and acceptance of autonomous shuttles among both operators and passengers. Addressing this issue requires innovative solutions that balance safety with operational efficiency in highly interactive urban environments.
Within the collaborative framework, Fraunhofer IESE contributed to three key aspects of the system development. The primary scientific contribution involves the integration of dynamic safety mechanisms to enhance the performance of the target system by employing realistic, human-like assumptions regarding pedestrian behavior at crossings during runtime. Initially, to ensure safety, the system adopts a static, worst-case scenario parameterization of pedestrian behavior, which unrealistically assumes that pedestrians could suddenly move at speeds comparable to Those of an Olympic gold-medalist. This conservative approach does not align with typical expectations of human drivers, who assess pedestrian behavior more dynamically.
Fraunhofer IESE helped to enhance the system by analyzing and modeling pedestrian behavior from a human driver's perspective, thereby enabling the system to adjust safety parameters dynamically to better match real-time situations. This approach makes the autonomous system's responses more intuitive and human-like, which in turn enhances acceptance among operators and passengers as the system’s behavior resembles that of a human driver.
The significant enhancement in system performance achieved by utilizing safe yet more realistic safety parameters for specific situations is accomplished through the "Situation-Aware Dynamic Risk Assessment" (SINADRA) method developed by Fraunhofer IESE. This method is complemented by Hitachi, Ltd.'s "Symbiotic Safety" approach. Together, they provide a robust framework for improving the interaction between autonomous vehicles and pedestrians in dynamic environments.
A secondary contribution centers on the adaptation of Safety of the Intended Functionality (SOTIF). Fraunhofer IESE’s own "Conditional Safety Certificates" (ConSerts) technology empowers the system to dynamically adapt to SOTIF faults emerging at runtime. This capability ensures that the system can accurately and safely assess its own operational capabilities. Furthermore, any capabilities essential for the implementation of other safety technologies in this collaboration that are impacted are addressed through system adaptation. This process facilitates a robust and reliable application of dynamic runtime risk management strategies by allowing appropriate system adaptations.
Lastly, with the help of the Digital Dependability Identities (DDI) technology, these models were transformed into software executables and integrated into the target system architecture of the autonomous shuttle bus and the BRT system.
The overall system including the different technologies was successfully evaluated for pedestrian crossing situations in simulations as well as on a real system in a closed testing area at Hitachi Campus, Japan. The evaluation showed that system performance improved, so that now the system does not need to wait at a pedestrian crossing every time it encounters a pedestrian, as illustrated in the following figure:
That's what Dr. Masaya Itoh, Autonomous Control Research Department says.
The integration of various dynamic runtime safety mechanisms –including Hitachi, Ltd.'s Symbiotic Safety, alongside Fraunhofer IESE's SINADRA and ConSert technologies – was implemented in a real system along with newly developed systematic approaches for dynamic safety. The effectiveness of these technologies was demonstrated through a real-life evaluation of the developed system. The next phase involves expanding the autonomous shuttle system to cover a broader range of scenarios, with the objective of deploying it on the public BRT road.
Additionally, Fraunhofer IESE’s reference process for dynamic safety engineering was utilized in the BRT shuttle use case. This process aids in maintaining traceability of different artifacts and supports robust safety argumentation. Leveraging this expertise, Hitachi, Ltd. is now equipped to apply the dynamic safety engineering process in other use cases and future development projects across various domains.
Fraunhofer Institute for Experimental Software Engineering IESE