ARM is best known as the company that has created key processor designs that power tens of billions of devices around the world including virtually every smartphone and tablet. Most modern cars and trucks also contain multiple ARM-based compute devices, in some cases dozens of them. But ARM doesn’t actually build and sell chips, they just license designs and instruction sets to other chip companies. Now ARM is leading an effort to create standardized software frameworks that make it easier for the automotive industry to develop applications that run on ARM devices.
One of the big trends in automotive over the past several years has been the shift toward the so-called software defined car. Tesla
Much of the industry sees this as an opportunity to create new revenue streams through feature subscriptions in addition to improving the customer experience. Whether customers actually buy into subscribing to features remains to be determined, but OTA will certainly be a pathway to providing a range of vehicle updates. One of the big challenges for automakers in going down this path is managing the software development and deployment process across their entire model range and over multiple model years.
Over time, as the underlying compute platforms evolve and improve, manufacturers would like to adopt the latest and greatest technologies. But until now, they have tended to stick with known quantities as long as possible to minimize the variations they have to support.
That’s where ARM’s Scalable Open Architecture for Embedded Edge (SOAFEE) comes in. SOAFEE actually includes two components, a software framework and a hardware reference platform. At a high level, the software framework is an abstraction layer that separates the applications such as the driver assist software, infotainment, body controls or other features from the underlying compute hardware.
Analogous examples would be Microsoft’s
In the case of automotive, many of the processors running in today’s vehicles contain ARM designed cores. These processors are supplied by companies such as NXP, Renesas, Nvidia and others. However, each of these companies licenses CPU core designs or other components from ARM and mixes them with a range of elements. For example, Nvidia’s Xavier chip that is now being used by several automakers for driver assistance systems contains 8 ARM CPU cores, an Nvidia Volta GPU, input/output ports (IO), memory and additional artificial intelligence accelerator logic. Chips from other suppliers combine a variety of other components on the silicon.
The initial release of SOAFEE is designed to work with a reference hardware platform that can now be preordered from Adlink that uses a 32-core Ampere Altra system on a chip (SoC) and a variety of IO. A ruggedized version for in-vehicle use with an 80-core SoC is also available.
Automakers can use these compute platforms for initial development of applications through SOAFEE. As other ARM licensees and partners produce their own variants of SOAFEE that target their specific hardware platforms, the goal is that automakers can simply transfer their applications over without reworking them. Over time, as an automaker has several years of models that may have silicon from different vendors, they should be able to update applications once and deploy to multiple platforms.
The SOAFEE framework is designed to support cloud-based containerized applications as well for development and deployment. Among the partners that have worked with ARM in defining SOAFEE are Amazon Web Services, as well as software vendors such as Green Hills, Red Hat
If the industry adopts SOAFEE broadly, this could lead to significant cost savings and faster deployment of updates across a broader range of vehicles. As automakers adopt new generation electrical/electronic architectures that shift from the current model of dozens of discrete electronic control units to a more centralized approach, SOAFEE could be an important enabler to make the transition smoother. More information about SOAFEE and the initial reference software can be found at https://gitlab.arm.com/soafee.