Automated, Continuous and Systematic Identification of System Issues

Proving functional safety is one of the key challenges for the market launch of automated driving. Traditional release methods based on kilometer-based test drives are not applicable for such systems, as they cannot be performed under economically reasonable conditions. Here, we present an alternative AI-based approach to generate critical scenarios for driving functions in a fully automated manner. The test cases found are then made available to developers and testers for further analysis and improvement of the system under test. Using an established simulation environment (e.g. CarMaker from IPG Automotive), the AI learns weak points of the driving functions in early phases of development and points them out to the function developers for continuous improvement.

RevoAI's AI-based environment was developed specifically for this application and enables the target-oriented generation of new critical scenarios in the simulation with a high degree of parallelization. For example, integrated with CarMaker on an HPC system, many thousands of scenarios can be efficiently generated and simulated overnight, and are readily available for developers on the next day. In benchmark studies, it can be shown that the number of critical scenarios that can be found with this method is significantly higher than with conventional methods.