The development of dSPACE Automotive Simulation Models, or ASM for short, began two decades ago. The tool suite for the simulation of combustion engines, vehicle dynamics, electrical components, and traffic environments already has a long history of success behind it, explains Dr. Herbert Schütte, who initiated the ASM development and has been driving its further development for a long time. Michael Peperhowe is now taking over the further development of ASM as Director Simulation Models & Scenarios. The future agenda includes topics such as the integration of AI and cloud technologies, functional enhancements for drive variants and AD applications, and increased support for the area of digital homologation.
How did the first version of ASM come about 20 years ago?
Dr. Herbert Schütte: At the time, we had a quasi-exclusive collaboration with a model manufacturer that had already begun in the mid-1990s as part of customer projects. This collaboration was very successful for many years. Later, the strong growth of HIL, the increasing functional requirements, and the globalization of our business presented us with challenges that we could only solve with our own modeling expertise.
Of course, we knew that we needed clear unique selling points in order to gain a sustainable foothold in this occupied and know-how-driven market. We had already received a lot of feedback from our HIL customers. For example, customers complained that they were unable to view implementation details from competitors for IP protection reasons. This led to a lack of confidence in the quality of the models and made it difficult for users to add their own model parts or replace partial models.
Another key point was the ease of use of the models and, in particular, the support for parameterization of the models. Compared to established providers, we attached great importance from the outset to an intuitive graphical interface for parameterization, extensive import and export filters, and support for data input from measurements.
How have the ASMs been functionally developed?
Dr. Herbert Schütte: We started with driving dynamics and gasoline/diesel models. This was later followed by road and environment models, an easy-to-use maneuver control system, and engine models that could also depict the combustion process and thus the internal cylinder pressure. The aim was to close the loop between the rapidly developing vehicle sensor and actuator technology in real time, in order to be able to test the latest control units in the vehicle. We have also continuously developed the hardware interfaces in order to be able to simulate linear lambda sensors or exhaust gas aftertreatment systems cleanly, for example. In recent years, the focus has often been on electrification in vehicles and the testing of ADAS/AD systems.
What role does ASM play in the automotive industry today and how has it influenced the industry?
Dr. Herbert Schütte: Despite our relatively late market entry in 2005, ASM is now an established player and is used by our customers worldwide, not only on HIL simulators. Some customers have standardized on ASM in order to leverage synergies and align their internal processes. Another important success factor is certainly the fact that we have modeling experts worldwide who can support our customers in the use of ASM.
Michael, you wrote your thesis on ASM models for combustion engines 20 years ago. An incredible amount has happened in the field of drive systems with electrification in two decades. Now you are taking over responsibility for further development. How is dSPACE positioned today with models for different drive types??
Michael Peperhowe: In the past, there were essentially the two combustion variants with auto-ignition and spark ignition. Today, electrification and synthetic fuels are leading to a significantly greater diversity of variants. This requires numerous extensions in the simulation: from various hybrid drive types to pure BEVs with different arrangements of the engine units to hydrogen burners or even fuel cells. The entire range of diverse drives can now be implemented with the expanded ASM portfolio, which surprises and ultimately convinces many customers with regard to the breadth and quality of our simulation models.
What are the special challenges in validating autonomous vehicles compared to simple driver assistance systems and to what extent does this affect dSPACE's simulation offering?
Michael Peperhowe: The main difference between ADAS (SAE Levels 1 and 2) and AD (from SAE Level 3) is that the responsibility for driving the vehicle is transferred from the driver to the vehicle and therefore to the manufacturer. This is associated with significantly higher requirements for validation and the consideration of a very large number of different sensors of various types such as cameras, radar, and lidar. For the field of simulation, this means that in addition to detailed driving physics, traffic flow simulations and physics-based sensor models must also be integrated into the application. One challenge is to feed the various sensor information into the different control units synchronously and in real time, i.e., at a predefined time interval. Based on the ASM Tool Suite, we have unique solutions on the market that make reproducible tests on the HIL possible in the first place.
Speaking of testing: Test coverage according to the operational design domain (ODD) plays a particularly important role in the validation of autonomous driving functions.
With the ASM Traffic module, we have created a way to run scenario-based tests in the cloud in a highly parallel manner. This means that thousands of test results can be generated in a very short time, especially in the critical areas known as corner cases.
How will modern technologies such as AI be integrated?
Michael Peperhowe: The buzzword AI has naturally also become indispensable at dSPACE. We even have an independent AI team that deals with various issues in this field. Among other things, we are working on the use of large language models (LLMs) for the textual definition of driving scenarios including the underlying road network. To this end, we have entered into partnerships with various cloud providers, as an appropriate cloud infrastructure with corresponding services is essential for such tasks.
In addition, we are receiving increasing requests from customers to test or even train their AI-based algorithms in the field of object detection with camera and radar systems. This requires an enormous amount of training data of the appropriate quality. In the field of perception in particular, this requires 3D worlds in the simulation that are no longer created manually, but are generated automatically using procedural methods or generative AI. The big advantage of virtual training data is that the reference, the so-called ground truth, is supplied free of charge, enabling a pixel-perfect evaluation of the actual perception. We are also constantly expanding our range of simulations in this area.
What future developments and innovations is dSPACE planning for ASM?
Michael Peperhowe: The topic of simulation will play an increasingly significant role in the homologation of driving functions. The real test drives are significantly enhanced by virtual simulation drives. This results in a wide range of requirements for the underlying simulation models, particularly in terms of reliability and quality. For this reason, we had ASM certified to ISO 26262 two years ago and we successfully undergo recertification every year. The next question is how well the simulation results can be trusted in terms of realism, the so-called trustworthiness or model credibility. Here we have already achieved very good results in the validation of the vehicle dynamics simulation, i.e., we can say with confidence that the ASM Vehicle Dynamics Simulation corresponds very well with real measurements. The next step is to validate the sensor models, which can also be achieved thanks to a large partner network with all well-known sensor manufacturers.
In addition to supporting the digital homologation described above, we will also focus on shaping ASM in the direction of standards. For example, we are currently introducing a traffic flow simulation that is natively based on the ASAM OpenScenario standard. This means that lossy conversions from and to the standard are a thing of the past.
We will also continue to open up ASM to other platforms. Even though a turnkey solution clearly offers functional and performance advantages, we see the need for customers to integrate various ASM modules into existing ecosystems. One solution for this is the provision of ASM using the FMI (Functional Mock-up Interface) standard.
And last but not least, we will continue to expand the ASM range in line with the wishes and needs of our customers and the market.
So, stay tuned.
