Last Fall, Arm announced the formation of the Autonomous Vehicle Computing Consortium (AVCC). The group of leaders within the automotive and technology spaces have a collective goal of accelerating the delivery of safe and affordable autonomous vehicles. While it’s a lofty goal, especially the latter, it’s great that someone (or group of someones) has put a stake in the ground to attempt to make this happen.

One of the early initiatives of the group is to define a reference architecture that will allow all parties to move forward on design and manufacture. This includes everything from power and thermal issues to requirements for software APIs for each building block within the autonomous system.

Another Consortium?

My first reaction back at the time of the announcement was simply, “Is there a need for another consortium?”

I asked that question of Chet Babla, Vice President of Arm’s automotive team. His response: “I recognize that there's a danger of people getting fatigued from too many consortia, but we do believe this has a valid existence. It's addressing a specific need and challenge for the market.”

Since the initial announcement about four months ago, working groups have been and continue to be formed. There are now working groups for architecture and imaging, and a few more that are near announcement.

Babla noted that consortia, by their nature, tend to move slowly because of the number of people, organizations, and ideas that are involved. He added, “Now, we’re starting to see some very good contributions being made by different member companies and technical discussions are starting to open.”

While lots of information related to the AVCC is open and public, there’s just as much that sits behind the curtain for the time being. But there is a stated goal to publish open guidelines around a “common compute platform” before the end of this year.

Deciding Before You Decide

Another key component has to do with the guidelines on the common compute platform, which are the building blocks, rather than the actual specs themselves. In other words, it’s important to define what’s needed in the specification before actually defining the spec itself. And that’s more of what’s happening today.

The original goal of accelerating the delivery of safe and cost-effective autonomous vehicles at scale remains intact. “That’s still a big looming challenge for the industry,” says Babla. “We're seeing that a lot of the hype around autonomous vehicles is giving way to reality, and people are realizing what a difficult challenge it is. So, the consortium is a pooling of resources across the industry to understand how we solve this compute challenge, particularly in a way that’s thermal- and power-efficient and cost-effective enough to be deployed into commercial vehicles.”

Ensuring Implementation Agnosticism

A key element of the AVCC is to be implementation agnostic. It’s likely that, as the working groups dive into the details, they’ll learn more about how the different groups must interact. Good examples are security and functional safety, which really need to be baked into each element of the design.

What’s happening is that as the technical working groups form and have these detailed ongoing discussions, they’re finding the need for additional working groups. And that will likely continue as development continues.

Babla hopes to see a rapid growth in AVCC membership this year. “We've already got a great selection of tier-one silicon players and OEMs. I'd like to see more because there’s strength in numbers.”

Why Arm?

One question that deserves an answer is why would Arm be taking a leadership position in the consortium, rather than have it led by one of its partners, one of the Tier One suppliers to the auto industry, or one of the auto makers themselves?

Said Babla, “We are constantly in touch with everyone in the design chain, including the Tier Ones, the OEMs, the software ecosystem, and even the government regulators. So it makes sense for Arm to have a leadership role here.”

But that’s not to say that Arm is working independently by any means. Many of the members play prominent roles. That includes the likes of Bosch, General Motors, Renesas, Synopsys, Nvidia, and NXP. And there’s likely more to come, as the technology advances and more companies get involved.