Digital Twins In Automotive

The term “digital twin” refers to a new principle that is gaining importance in the development of complex hardware/software systems. In general, it refers to a virtual representation of the real system. This model serves to simulate the functional interactions of the parts, saving time and money by avoiding unnecessary redesign cycles and enabling considerably better optimization of the overall system design.

While still relatively new to the area of automotive development, the digital twin concept has been used for years in the field of industrial automation. The system modeling approach already is in widespread use in automotive, particularly with the goal of verifying the system behavior, but the digital twin represents a significant addition.

Digital twins typically have a longer lifetime and greater efficiency compared with a classic system model since they can be re-used at multiple points along the entire product lifecycle. Use of the digital twin as a virtual prototype typically already starts in the planning phase for the product development, allowing more effective exploration of the design space and supporting architectural decisions.

In the following step, the digital twin serves as a communication tool at the interface for system implementation. In the form of an executable specification, the resulting model can be shared with internal developer teams and external contractors. It acts as the gold standard reference for the implementation phase, ensuring clear communication and avoiding errors due to misunderstanding of the specification. In the ideal case, the tests for checking compliance with the specification values are provided at the same time to ensure a consistent test concept from the very start.

These digital mock-ups also can be sent to customers for integration into a larger system landscape. IP remains protected thanks to an abstract model, which does not contain the details of the actual implementation. For example, software developers can receive virtual prototypes of hardware that is either still in development or should not yet be shared in detail. These familiar system development approaches can be implemented very easily with a digital twin.

Finally, the digital twin remains useful even beyond the development phase. Data from operation of the real twin in the field can be played back on the virtual twin to improve quality of the model, and future development work on the product will profit from a more accurate representation of reality.

Some of the approaches described above already are employed in the development of automotive systems. Due to the many safety-critical systems in a vehicle, a number of binding requirements apply to the development process for automotive electronics. The digital twin can help guarantee that these requirements are met. For example, the responses of the system to possible faults must be investigated and evaluated in advance. A virtual prototype is perfectly suited for injecting and simulating faults.

The major providers of commercial development environments for integrated electronic systems are working intensively on approaches to virtual prototyping. One area of particular note here is chip development, where emulator solutions and high-performance FPGA platforms are offered. However, the digital twin approach captures more relationships at the system level. Another important aspect is the interaction of electronic components with multi-physical effects, such as in the case of mechanical components or changing thermal conditions.

Such a complex digital twin must be capable of integrating sub-models of very different origins. This requires a powerful, open interface, such as provided by an FMI (Functional Mockup Interface), which permits the integration of models based on completely different mathematical principles, description languages and abstraction levels.

One future challenge lies in expanding the digital twin from its initial functional focus to include non-functional properties. This could allow optimization of the power consumption of the complete system in various operating models, for example, as well as open up the possibility of investigations of safety and reliability. The sheer number of test kilometers required to ensure the safety of highly complex assistance systems and automated driving functions will be enough to make virtual development with digital twins indispensable in the future.