Investigation of the Degradable Safety architecture in the AI environment

Investigation of the Degradable Safety architecture in the AI environment

연구 분야: Verification

학회: ICCAI '24: Proceedings of the 2024 10th International Conference on Computing and Artificial Intelligence

The approach in this research has focused on a specific environment, knowing that the results can be adapted to numerous other areas with very little change. The new and unique feature is a security architecture (4oo6) which is "generically" safe in sub-applications (e.g. the interface applications). This means that in the future, applications can be set up quickly and inexpensively on low-cost hardware with little to moderate effort. This also enables the implementation of new products with high functional safety requirements. Since there are no complete certified safety concepts for autonomous driving or systems on the market yet, the need for such safety considerations and results is very high. The work therefore also had the goal of preparing all important safety and process steps for the purpose of series development of autonomous systems up to safety certification. Therefore, the research investigated new process approaches and error possibilities, which will greatly simplify auditing or certification in the future.