Not that long ago, prospective car buyers flocked to auto shows to gape at the latest models and compare the specs that mattered: horsepower, engine cylinders, and how fast they could accelerate from zero to 60 miles per hour. Today’s discerning shoppers care more about onboard software and advanced driver safety features, such as collision warnings and emergency braking. Tomorrow’s showroom visitors, however, are less likely to care about the act of driving than its opposite—everything else their cars can do on the way from A to B. To that end, automakers, technology firms, network operators, insurers, and media companies (to name but a few) are racing to rewire automobiles into what amounts to computers on wheels. At stake is a captive audience of literally billions and new markets in hardware, software, entertainment, and more. The research firm Strategy Analytics, for example, predicts this emerging “passenger economy” of passively enjoyed goods and services will total $7 trillion by 2050.
Providing these services will require data—a lot of data. According to a McKinsey & Company estimate, connected cars create up to 25GB of data per hour, which equates to dozens of movies stored in HD every 60 minutes. Collecting, storing, processing, analyzing, and sharing that much data will require a corresponding transformation in sensors, mapping, displays, on-board computing, wireless networks, telematics, and the standards undergirding all of this.
As with previous computing platforms, building connected and autonomous vehicles (AV) will require multi-industry collaboration and competition. One of these firms is Harman International, the connected device maker acquired by Samsung in 2017. FastCo Works spoke with Vishnu Sundaram, vice president of Harman’s telematics business unit, about the challenges of creating a car that will make you forget about driving.
As a technology firm designing hardware and experiences inside vehicles for the automakers—your customers—what is Harman’s vision for how connected and autonomous vehicles will evolve? What exists today? What’s missing?
Vishnu Sundaram: The vehicles of the future will be differentiated by the intelligence inside them, rather than features like horsepower or better suspensions. To that end, access to information in the cloud will be fundamental, so connectivity will be the backbone of these cars.
I think 5G is around the corner, and offers several improvements over 4G, especially low latency, high bandwidth, and extended range. I call it the “1, 10, and 1″—one millisecond of latency, 10 gigabytes of bandwidth, and a one-kilometer radius—which together will enable new use cases around safety. All of the OEMs [original-equipment manufacturers] plan to introduce this between 2022 and 2023.
After that are autonomous cars. We already have driver-assistance systems on the market now, but the cars will need to rely on as many sensors as possible to achieve hands-off driving. That means going beyond the sensors installed in the car to accessing the sensors of all the cars around them—something that’s possible due to the low latency of 5G. They also need live traffic conditions and mapping, which is something that can be delivered over high-bandwidth links.
A lot of sensors means a lot of data, so the cars will also need an infinite amount of storage and computation. This can’t be contained within the car, obviously, so we believe the combination of 5G and edge computing will be essential in transitioning from today’s to future cars.
What do you mean by an “infinite” amount of computation? For safety reasons, aren’t automakers designing on-board guidance and navigation systems that can operate independently, just in case? Where does the cloud meet the car?
VS: If you look at the evolution of autonomous vehicles, it started with a fragmented technical architecture, with multiple on-board pieces. It’s now moving toward a centralized architecture, and the future we see is a distributed one, where part of the computation is in the car, and part of it in the cloud and on the edge. This will be a paradigm shift.
It won’t eliminate the minimum fallback computing that’s needed for safety. But when the car has access to the edge, it’ll be able to derive additional intelligence from the same sensor data to make better decisions, or to deliver high-definition media by downloading it from the cloud.
This will also have an impact in the ecosystem. For example, 5G needs a lot of support from the mobile operators, starting with investments in network infrastructure. Mobile operators expect high-value services offered on top of 5G, one example of which will be edge computing.
What role do standards have to play, and how do we avoid having the kind of battles that plagued the early days of other essential technologies such as PCs and mobile phones?
VS: For vehicle-to-vehicle (V2V) communication, there are fundamentally two choices: DSRC (dedicated short-range communications) and C-V2X (cellular vehicle-to-everything). If the decision had been made a few years ago, it would have been DSRC, but the window has moved on to CV2X, which is fundamentally part of 5G and so would require one less new network to be built on the ground. With the passing of time, it should exceed the performance of DSRC.
That said, some regions have already made decisions. China is clearly moving in the direction of CV2X, while Europe and the U.S. are still choosing between them. This is of interest to us, because it could create a regional divide. The downside is that everyone would need to invest in multiple systems.
Another hurdle is that standards for interoperability haven’t been defined yet. What this will probably lead to is OEMs sharing sensor data between only their own vehicles. That’s step one. Step two will be OEMs collaborating with only a few others, and step three will be eventual consolidation around one standard.
But one thing I can tell you is that it’s all a game of time. The FCC will choose to deploy one technology that will be standardized. The debate will only persist for another year or two.
How will vehicles evolve once 5G is mainstream and AVs have evolved to a point where cars are effectively rolling entertainment systems?
VS: Once high-speed networks are available in the cars, we’ll be able to bring shared content to the passengers. The immediate, dramatic change will be an increase in the number of displays. We see them replacing traditional controls, and that’s just the start. As we move forward, we envision a lot of new technologies will appear for enhancing the user experience—we’re already working on a headset display that can project inside an AV. And we’re also looking at AR/VR-capable systems that would really push the envelope in terms of what’s available.
How will connected cars accelerate the development of autonomous ones?
VS: AVs need to be constantly learning, because there’s a lot of real-world training data needed to perfect them. Building closed-loop systems that rely on learning from one car, relaying that insight back to the cloud and then [downloading] it to all the other cars is necessary. These systems don’t exist yet, but most OEMs are planning to deploy one.
Today, they’re using an offline approach, where they download all the data into a hard drive and when a car goes in for service, they’re given the latest update. This will continue until we have ubiquitous 5G coverage and a new business model, more or less, for OEMs to consume a lot more bandwidth from carriers. Once that happens, AVs will start learning much faster.
This article was created for and commissioned by Harman.