Efficient Failure Rate Estimation Using AI-Based Simulation of Critical Scenarios

The article delves into the application of AI-based simulation for estimating failure rates in advanced driver assistance systems and automated driving functions. By leveraging AI-driven parameter space exploration and the simulation environment CarMaker, the approach identifies critical scenarios and corner cases efficiently. This method not only minimizes failure rates but also ensures continuous improvement by alerting function developers to potential weaknesses early in the development process. The article highlights the use of various data sources, such as fleet recordings, infrastructure sensors, and drone data, to extrapolate ground truth information. It also emphasizes the importance of matching real-world parameter distributions in simulation to accurately estimate failure rates. The AI-based sampling methods developed by RevoAI are showcased for their ability to determine failure rates quickly and reliably, significantly outperforming conventional methods like Monte Carlo simulations. The article further discusses the scalability of simulations through parallelization and the validity of simulations based on high-quality components and data sources. Overall, the article presents a comprehensive and innovative approach to enhancing the safety and reliability of advanced driving functions through efficient and effective simulation techniques.