Some of us still cherish the memories of (somewhat uncomfortable, yet delightful) journeys with the famed Trabant 601. But the class of vehicles that this legendary model pertains to, those containing a bare minimum of electrical and electronic components, is gone forever. Digitalization has transformed the mobility industry and is changing the fundamentals of automotive engineering.
Today, as cars are becoming networked supercomputers on wheels, the architectures they are built on are in dire need of radical modernization. This is no simple facelift: leading automotive engineering experts today are designing the next generation of automotive electronics architectures that will change how cars are developed for decades to come.
Architectural challenges of digital mobility
With digitalization in the mobility industry, electrical and electronic (E/E) components have grown to become a primary focus of OEMs and automotive suppliers worldwide. Advancing E/E technology has contributed to the advent and skyrocketing role of software-enabled features in modern cars, affecting everything from infotainment systems and general user experience to intelligent safety features, and more.
As assisted driving and Highly Automated Driving (HAD) technologies pave the way to full autonomy, the complexity of software built into road vehicles today has exploded and continues to grow exponentially. Add to this the incredible amounts of sensor data and the challenges of sensor fusion in modern automobiles, and the need to manage, safeguard, and make sense of all this data.
This multiplication in the amount of features, data analysis, and driver assist capabilities delivered by software necessitates the immense growth of computing power. With more software, and the trend of “appification” in automotive technology comes the requirement of over-the-air (OTA) updates. Remote updates enable optimization and will help guarantee safety along the vehicle’s designed lifecycle.
Therefore, today’s architecture of functionally isolated control units is reaching its limits. The partitioning of functions based on the availability of computing power, and the heterogenous development methods used today to deliver these capabilities can no longer be sustained. This fast-paced evolution of automotive electronics has introduced a need for a new architecture that the development of modern cars will rest upon.
Modularized SOA and generalized computing platforms
The automotive domain today is adopting and adapting solutions from consumer electronics, drawing on cutting-edge advancements in digital technology. The way forward in automotive technology is the consolidation of electronic control units, giving way to a modularized Service-oriented Architecture (SOA).
Leading engineering experts at automotive innovators like BMW are working on a hierarchical E/E architecture based on a layered classification of electronic control units (ECUs). Consolidated layers (stacks) will serve the different requirements of specific vehicle functions.
What this means, in simple terms, is that components serving certain pieces of functionality will be encapsulated in layers on top of each other. For instance, there might be a high-performance layer for driver assist (ADAS or HAD) functionality, and a super fast stack optimized for low latency times required by the fundamental safety features of the car. The three main layers differentiated by the Automotive Open System Architecture (AUTOSAR) and BMW’s concept of hierarchical automotive SOA make the idea more tangible:
- A basic layer for services such as processing the data coming from the various sensors installed in the car.
- An extended middleware layer for inter- and intra-ECU data exchange. This layer is responsible for sensor data fusion including localization and the interpretation of sensor data (helping the car “understand” its environment).
- An application layer that uses all this data to operate certain features such as the brake assistant or the side-view assistant.
Image source: Future Automotive Architecture and the Impact of IT Trends, IEEE Software, 2017 vol. 34 Issue No. 03 - May-Jun, accessed on computer.org
Benefits of a Service-Oriented Architecture
This consolidation of ECUs and a layered model of modularized architecture offers many advantages for automotive developers.
For one, encapsulation and layering enable greater scalability, preparing the architecture for future challenges introduced by rapid innovation in automotive technology. Taking the SOA approach also reduces system complexity and facilitates the use of generalized computing platforms. This opens up opportunities for system-level optimization and facilitates the reuse of software components across vehicle variants and multiple generations.
It also enables developers to use unified development methods across layers including Agile practices, and robust testing against interfaces.
This uniformization greatly reduces the fragmentation in terms of design principles, processes, and tooling seen in today’s automotive development environments. With this next-generation architecture, the use of Scrum and shared code repositories is greatly facilitated, and the distributed development of requirements-based ECU classes becomes easier too. Developers can more easily apply Continuous Integration and early integration testing via virtual integration.
Overall, automotive developers benefit greatly from taking a smart approach to adopting and adapting methods used in the development of digital consumer technology. Applying Service-Oriented Architectures in automotive engineering reduces complexity, enables the use of Agile, and helps homogenize a currently fragmented development environment. All the while, this new approach helps carmakers maintain the highest standards of safety and reliability, IT security, product quality, and usability in automotive applications.
Future Automotive Architecture and the Impact of IT Trends by Matthias Traub, Alexander Maier, and Kai L. Barbehön
Rethinking car software and electronics architecture by Ondrej Burkacky, Johannes Deichmann, Georg Doll, and Christian Knochenhauer