Quality assurance is an essential part of the development process of automated driving systems. Therefore, it should be integrated into the development process as early as possible.
Scenario-based testing is a suitable method to complement the route-based approach of real drives and thus reduce the required mileage. Meanwhile, EU directives already require scenario-based testing in the approval process. Since there are many different focal points and methods of scenario extraction and evaluation, research is being conducted at FZI/ITIV on different approaches to scenario-based testing.
Scenario evaluation and criticality metrics
Most criticality metrics evaluate the situation of the first-person vehicle with respect to another road user. In the vehicle on the road and in simple scenario evaluations, this approach is sufficient, but required scenarios become more and more complex and datasets to be evaluated can contain up to 60 or 70 vehicles in a single scene. At FZI/ITIV, we are researching new possibilities and methods to provide an all-encompassing statement about complex scenes and situations and to leave no aspects of them unevaluated.
Scenario-based testing is a concept that is gaining recognition for automated driving testing. However, it raises the problem of which scenarios are relevant for testing and which are not. Using different optimization algorithms and evaluation metrics, critical and interesting scenarios should be identified, generated and combined to obtain a scenario catalog that is as comprehensive as possible.