Automated Driving

Validation of AI-Based Driving Functions

Processes and methods for the product validation of driving functions are not suitable for functions with learning (AI-based) algorithms. In particular, AI-based driving functions (e.g. functions using machine learning technique) are important for automated driving. In order to expand the currently practiced validation processes, the AQI has identified gaps in these used processes and methods. Accordingly, recommendations for action were formulated. Furthermore, the AQI is currently developing concrete measures on how the latest used validation processes can be adapted or supplemented by new approaches (especially regarding the VDA volumes). To continue this work, the AQI is formulating proposals for the extension of existing quality standards and norms to assure learning driving functions.

Approval and Operation of Simulation Tools for the Validation of Automated Driving Functions

Due to the large number and complexity of test cases, it is not possible to proof the correctness of automated driving functions in field tests only. Instead, an additional validation by simulations is to be integrated into the development process. While activities for specifying and defining the digital/simulated test scenarios are ongoing, there are no consistent procedures for the approval and operation of the simulation tools used for the validation. The AQI has developed recommendations for action and proposed solutions to establish the development of standardised procedures for the approval and operation of these simulation tools.