Before highly automated and networked vehicles are allowed on German and European roads, their safety must be comprehensively proven. While conventional vehicles are subject to established and officially recognised certification methods and regular testing, the rules for the validation of automated driving systems, in which the driver relinquishes control for longer periods or on certain sections of the route, are still in their infancy. Simulations, or digital tests, will play a major role in developing these rules. In the simulation-based development and testing of automated driving (SET Level) project, the German Aerospace Center (Deutsches Zentrum für Luft- und Raumfahrt; DLR) Institute of Transportation Systems is working with 19 partners from industry and research to develop the necessary digital tools and to link them together in a 'tool chain'. These tools will simulate as many traffic situations as possible as reliably and in as much detail as possible. By doing so, a significant proportion of physical driving tests can be transferred to simulations. This saves time and costs and makes the certification process more efficient and comprehensible.
Simulated tests as a prerequisite for the certification of automated vehicles
For an automated and networked vehicle to be suitable for the reality of road traffic, it must successfully master several thousand traffic situations. Testing them all in practical road tests is beyond the scope of any certification process. "Only with powerful simulation-based tools and methods which can be used during development and during the certification process can automated vehicles be safely brought onto the road and into use," explains DLR's Frank Köster, one of the SET Level project's two coordinators. "Simulations are an important addition to existing techniques, such as test stands and test fields." DLR's primary contribution to the project is its many years of experience in the digital and physical operation of test fields and test infrastructure, as well as its expertise in the field of digital twins. Digital twinning is a technique used during technology development for transferring objects or processes from the real to the digital world. In addition to the DLR Institute of Transportation Systems, two new DLR institutes currently being established, the Institute of Systems Engineering for Future Mobility and the Institute for Artificial Intelligence Safety and Security, also conduct work in this field.
Converting complex realistic scenarios into code as efficiently as possible
The SET Level team face a host of challenges. In addition to the car that will be tested – including its software, sensors and control technology – they must also create a digital image of the complex surrounding traffic environment. This includes roads, infrastructure, vehicles of all kinds, other road users on foot or on bicycles, different weather conditions and disruptive factors. The selection of test scenarios is also key. "Even using simulations, we cannot cover all theoretically conceivable scenarios," says Hardi Hungar, scientific lead for the DLR activities in the project. "That would see us carrying out calculations forever. Instead, we have to reliably select those scenarios that are representative for the assessment of the vehicle's safety. Addressing the questions around how to arrive at these scenarios and how to comprehensively test them in the simulation is also an important part of the project." Hungar goes on to describe the simulated tests: "The virtual test vehicle drives through the replicated traffic area while situational assessment results and test outputs from the vehicle are displayed in parallel. This allows you to follow how the test vehicle deals with different challenges."
The project team has already reached its first milestone. It has developed the individual components of the simulation and demonstrated that these tools are compatible with one another. On 29 April 2021, the researchers presented these results to an international audience of experts at the project’s virtual mid-term event using three simulated traffic scenarios.
Another special feature of the SET Level project is that it relies as much as possible on open-source solutions. This allows the methods produced to be used and further developed by many companies and research institutions after the end of the project. It also has a noticeable effect on the nature of the project's partnership and cooperation. “Together we want to establish methods and set standards for the certification processes for highly automated vehicles in Germany and Europe," says Henning Mosebach, Head of Research Strategy at the Institute of Transport Systems. "The project's diverse consortium has come together to achieve this."