Go to content

SIEMENS

Research & Development
Technology Press and Innovation Communications

Dr. Ulrich Eberl
Herr Dr. Ulrich Eberl
  • Wittelsbacherplatz 2
  • 80333 Munich
  • Germany
Dr. Ulrich Eberl
Herr Florian Martini
  • Wittelsbacherplatz 2
  • 80333 Munich
  • Germany
pictures

Thanks to sensors, embedded computers, and communications technology, the car of tomorrow will increasingly become a robot on wheels.

In the context of the “RACE” research project, Siemens and its partners are examining a new software architecture for tomorrow’s vehicles.

The concept envisions a dramatic simplification from the multitude of control elements in today’s cars to the architecture of the future.

The growing number of software functions will be distributed over fewer hardware components.

Road to a Digital Future

More and more automotive systems are driven by software. Driver assistance systems and electric vehicles are accelerating this trend to such an extent that developers are envisioning a completely new architecture for tomorrow's cars. In a research project called "RACE," new concepts are already being developed and tested in prototypes.

Image In the context of the “RACE” research project, Siemens and its partners are examining a new software architecture for tomorrow’s vehicles.

Our cars are rolling computers. High-end automobiles contain up to 100 control units and dozens of sensors. Engines, transmissions, brakes, airbags, and even power windows are equipped with intelligent electronics, which can recognize, for example, when a child’s hand is between the window glass and the window frame. There are systems that help drivers stay in their lanes and park, warn them that they might be getting too tired to drive, and coordinate the wheels when driving on slippery surfaces.
Some types of cars are already equipped to brake automatically in critical situations. This action involves not only the electronic stability control system but also a kind of cerebellum in the transmission that automatically causes it to downshift while the passenger safety equipment pulls the seat belts tight. Since all of the control elements on board modern cars are heavily interdependent, the result can be a kind of anarchy that can put excessive stress on an automobile’s communication system, or “bus.” This is the component that makes it possible for subsystems to exchange data with one another.

As more and more functions become increasingly autonomous, it will become difficult for the current standard of decentralized intelligence — electronics and software modules distributed throughout a car — to perform adequately. Despite ever faster connections and protocols, there will eventually be a traffic jam in the data flow. “We have to deal with the causes of these problems and not just the symptoms,” explains Professor Gernot Spiegelberg, who is responsible for electromobility concepts at Siemens Corporate Technology.

Spiegelberg’s suggestion for a solution is based on the way the human brain works. In the same way that the brain has specific regions that are responsible for functions such as sight, motor control, and memory storage, so too must a central computer have a function-oriented software architecture that drives a car. This approach would allow data processing resources to be efficiently applied, as well as the rapid evaluation of complex traffic situations. Individual functions could be upgraded or replaced at any time, and only minimal effort would be needed to transfer the software package from one type of car to another.

As simple as this may sound, it would be a revolution for the auto industry. It would mean that every supplier of electronic parts would have to deliver associated software modules that could speak to one another through logically predefined interfaces.

Evolution and Revolution. How might the electronic architecture for tomorrow’s automobile look? That’s what’s being researched by the RACE project, which is being coordinated by Siemens — the acronym stands for “Robust and Reliant Automotive Computing Environment for Future eCars.” In addition to Siemens, automotive supplier TRW, service provider AVL, and five renowned academic institutes are also participating in the project, which is supported by the German Federal Ministry of Economics and Technology to the tune of around €10 million. By the end of 2014 the partners plan not only to have worked out a theoretical description of the hardware, software, system schematics, and sensor integration, but also to have built two prototypes.

The first prototype, known as “Evolution,” will primarily demonstrate the transition from today’s architectures to those of the future. The biggest challenge is the high cost of developing completely new software. As Professor Manfred Broy of the Technical University of Munich confirms, “It’s obvious that in the long run we’ll need a different system architecture for automobiles. But we also have to be clear about how much money has been invested in current vehicle software and how expensive it will be to rewrite it and partly redesign it.” The Evolution subproject is also intended to show that a function-oriented architecture makes significant savings possible over the long term. For example, if a component, including its software, has been certified in accordance with the strict ISO 26262 safety standard, then it should be transferable to another vehicle from the same manufacturer without modification or further testing — as is already the case in the aviation industry.

The second prototype, by contrast, will show what a “Revolution” could look like. Here, a completely redeveloped system architecture will be fully implemented in a vehicle. The goal is to ensure that the electric drive, brake system, and all other functions relevant to driving will work so well that the car could be licensed to operate on public roads. According to Spiegelberg, “We’ll see that cost savings will be possible here as well, as the entire vehicle architecture will be designed in a completely different way from today’s cars.” All in all, this will make it easier to implement completely new vehicle concepts.

For example, the prototype will be equipped with wheel hub motors on the rear axle. They will not only give the vehicle excellent acceleration but also be able to generate so much braking power that a braking system that uses ablative brake pads will only be necessary on the front wheels. That would cut costs. And it wouldn’t require two control units — one for the brakes and one for the motors — to deal with longitudinal dynamics. The concept car will also have steering “by wire,” which means that the steering column as a mechanical connection between the steering wheel and the front axle will be omitted.

The exact structure of the software will be determined over the course of the project, which started at the beginning of 2012. Broy, the software expert, expects that client-server structures, service-oriented architectures (SOA), and layered architectures will all play a big role. All three ways of structuring complex software boil down to the adoption of a rigid hierarchy. In a layered architecture, components no longer have equal authority, which means that components from higher layers may use the elements of lower layers, but not the other way around. In a client-server structure, the software is distributed among different hardware components, but there is a clear definition as to which components have what authority. With service-oriented architecture, the software is structured according to areas of responsibility.

This applies to hardware in an analogous way, as increasingly intelligent sensors are being installed in cars. This is because current and future assistance systems rely on vehicles being aware of their environment. Stereo cameras, laser, radar, and — for close range — ultrasound sensors, will give tomorrow’s cars a 360° view, not to mention the fact that they will be equipped with decentralized intelligence. These “nodes” in the vehicular autonomic nervous system will take over the job of signal processing, while the vehicle’s “brain” will be responsible for situation recognition and, when necessary, for taking action.

RACE is also researching the interesting question of what control loop configuration would allow intelligent sensors to be directly connected to intelligent actuators so that the vehicle’s brain would not need to take action in most cases, but would merely monitor situations. One situation where the brain might intervene would be modulated breaking.

Upgrades for Older Cars. But what about the driver? He or she would profit most from new software structures that would make it easier to integrate functions retroactively. Updates for the infotainment system are already standard today, but when a luxury automaker introduces a new assistance system for collision avoidance, it is only available in new cars. In the future, on the other hand, the new information architecture from Siemens would allow older models to be retrofitted with the new software.
To protect such an open system from hackers, RACE is also examining the security of the new architecture. But Broy doesn’t see any fundamental problems. “In principle, we already know how to design software updates that are secure,” he points out. “It means building in a firewall, the introduction of clear security requirements, and developing a general security concept for the systems in an automobile.”
The race is now on to create a future-oriented, sustainable electronics structure for cars.

Johannes Winterhagen