When we first meet with advanced automotive design teams about their intelligent vehicle plans, they often assume the conversation will remain focused on platform speeds and feeds, CPU and GPU clock rates, and data latencies and throughput. This isn’t surprising as automakers are always interested in discussing the possibilities of maximizing compute performance in future vehicles, and this is squarely in what most people think of as Intel’s traditional wheelhouse. But in order to design a truly intelligent car, we need to start the design conversation by covering all systems that require intelligence. Learn more in this podcast: How Automakers are Advancing Intelligent Systems.
So meetings often swerve off of the agenda once we start discussing things like how they plan to capture data generated by the vehicle in an effort to make it situationally aware, whether they are using LIDAR or ultrasound and cameras, etc. That usually leads to a discussion about the fidelity they need to deliver high-performance capabilities via different types of sensors. And then:
Call in the sensor team.
When we inquire how they plan to move the mountains of data quickly, often in real-time across the vehicle bus architectures, they realize this is once again a very different conversation than the one they expected. Now the conversation has shifted from CPU to sensors to moving data. Then I ask how they verify the data is correct and hasn’t been hacked. And what about spoofing? Artifact data?
Let’s get the security team in here.
Then, we ask questions about monitoring the driver, the passengers, the overall in-vehicle cabin environment, when to apply advanced driver assist systems (ADAS), and when to return control to the driver. How do they notify the driver? How do they know the driver is ready to receive control? I ask if they’ve looked at the usage model in the cabin.
Go get the user experience team.
It’s not atypical for a planned two-hour meeting to extend to six hours as the number of participants on their side grows from four to sixteen or more. On our side, we are learning an enormous amount about automotive systems as well. Discussions about sensors, security, data movement, sensor fusion, and driver alertness are all crucial for harnessing, moving, and utilizing data in connected cars. Because in the end, of course, what’s imperative is that intelligent vehicles develop situational awareness and are ultimately smart enough to actually use the data being collected.
We love brainstorming these types of ideas with car OEMs because that is where we break new ground for what’s possible and figure out ways to build systems that scale entry model to premium, as well as scale model year to model year. We don’t yet know what we don’t know, but one thing we all agree on is that this is a compute problem (that just so happens to be wrapped in sheet metal). When meetings go several hours longer than scheduled, I know we’re in the right place—with the right people, asking the right questions, and investigating new and better ways to solve industry challenges.
Have any stories from marathon design sessions that you’d care to share? Drop us a line.
Listen to the podcast: How Automakers are Advancing Intelligent Systems.