When self-driving cars actually reach the market is still not clear, but the automotive ecosystem is preparing for huge changes.
The promise of autonomous driving is a significant one, with far fewer fatalities from vehicle crashes — down from 30,000 annually — topping the list of benefits. Beyond that, autonomous driving also promises increased convenience and productivity and less troublesome commutes.
But autonomous driving also pushes automotive technology into uncharted areas. There is little to fall back on as a basis for making many critical decisions about safety, security, or version control, such as what happens if one car updates its software and another doesn’t. Until a decade ago, most car companies never even thought cars would be connected to anything. As such, the automotive ecosystem is now grappling with the varied impacts, trying to chart the way forward.
“The industry recognizes it is a profound change,” David Schutt, CEO of SAE International, said at SAE World Congress last month. “What’s going on right now with electric vehicles, and connected vehicles —it’s going to change the way our cars are designed and made. It going to change the way we commute, the way we drive, the way we own cars. When you start to add autonomous technology, connected with a new business model such as Uber or Lyft, it really is fundamentally changing. Instead of having to sell two cars to every family, you might be selling 0.2 cars to every family, but it’s going to be a very different kind of car.”
Chris Rowen, Cadence Fellow and CTO of the IP Group, both agrees and disagrees. “There are many dimensions to how autonomous vehicles are potentially disruptive. One that people talk about [often] is if you have an autonomous car then people will rent or lease or taxi instead of owning cars, and ownership of cars in the U.S. is going to collapse, and it leads to the end of the automobile industry as we know it. That’s one of the grander disruption schemes, but adding self-driving capability, adding autonomy is probably not going to change at least the American love affair with the car.”
While some of these ideas are revolutionary, the implementation will likely be evolutionary as new technologies are developed, tested and refined.
“There’s not going to be a certain point in time when we suddenly switch over to complete autonomy. I am of the gradualist school that says we will gradually get more and more self-driving features that allow people to relax more and more of the time. But the fact is that some of the scenarios that you have to be ready for are so obscure and so difficult that it will take a long time to gather the experience to allow people to be truly 100% level for autonomous and their driving scenarios. Having the human operator there as back up is going to remain part of it, and the experience of being in a car will continue to be important even if it will change somewhat from being the fighter pilot experience that some people may want it to be, to more of the entertainment center experience.”
And while the cars will be better behaved than they would have been in the absence of autonomous technology, there may be just as many cars on the road.
“It will change the experience, but this idea that the number of cars will collapse is probably—depending on how you look at it—either too good to be true, or too bad to be true,” Rowen said. “Certainly when you are sitting in traffic, you’d love to be able to say, ‘I wish there were half as many cars.’ If you’re trying to invest in the technology, you’d like to say, ‘No, I actually do want a fair number of target units to address.’”
Further, it seems clear at this point that there will be more categories of cars. “There will be individual cars, there will be Uber cars, and there will be cars that are built for convoying as in, your pod in a virtual bus, that makes things happen. There will be a lot of exploration of business models around this as the urban experience diverges more and more from the rural experience. Certainly there are lots of people who, when they are in a parking-constrained situation, would really rather not have their own car. Parking alone will drive shared vehicles in a significant way in the densest of urban areas.”
Along with changes in technology are new standards, which are struggling to keep pace.
“Standards play an interesting role here because typically what happened in the past is that innovation drove very fast, then it settled, and we would standardize,” said Jack Pokrzywa, director of global ground vehicle standards at SAE International. “Now we are standardizing along with the innovation, so it’s not necessarily a race between innovation and standards. But it’s a necessity to have some foundation because those cars are appearing on public roads now, and there are still a lot of questions about it. So standards are becoming a fundamental part of this whole effort.”
SAE is working with the automotive industry on a wide variety of standards to enable what is being called connected transportation — including terms and definitions, safety, security, interoperability, privacy, and vehicle and system performance requirements.
One of the places where standards make sense is where it is pre-competitive, said Schutt. “So the industry is having to really think through, either actively or by default, where is it going to be pre-competitive, where does it not make sense to differentiate between one OEM and the other, and to really get this adopted. The government’s role is to provide safety for their citizens, so that’s where this is really merging. The standards are not being forced to drive the regulations, but as there needs to be regulations, and standards help us get there.”
In fact, the U.S. government is paying very close attention in this area. John Carlin, assistant attorney general for national security in the U.S. Department of Justice, recognizes there are a lot of potential benefits for autonomous vehicles in terms of the safety of drivers. “It might reduce death and injury like nothing we’ve seen since the advent of seatbelts, and you’ve seen other things like braking technology cause a reduction in terms of loss of life or injury, so there are many potential positive benefits.”
But there are security risks that need to be addressed here, as well. “At the same time, we need to think through whether it presents the possibility of not just an attack on one particular automobile, but an attack that might disable hundreds or thousands, or hundreds of thousands of automobiles simultaneously—think through the ways in which that could occur before the technology is deployed.”
Carlin noted that the government wants to work together with industry now as it designs toward what that future will look like, “so the best minds from the regulator side of the house, but also the FBI, informed by the rest of the intelligence community, can work with them on figuring out what the threats are and try to help them prevent them by designing technology on the front end.”
The impacts of autonomous vehicles on technology are just as significant as the regulatory, social and standards challenges.
Given the stricter safety and longer product lifecycle requirements, the vetting of technology may be an effective governor on the pace of change. “While it’s very important technology, it’s not something that will change dramatically year-by-year, so it is going to take a good 10 years to get to the point of a high degree of autonomy for the systems,” Rowen observed. “Even then there will be incremental improvements that continue for the next couple of decades beyond that. It is a long transition rather than something that happens overnight, but the transformation will be important not only in the automotive space. Then you’ve got to look at trucking and scooters and bicycles. What is the equivalent advanced driver assistance system for bicycles? You can imagine that you can make bicycles a whole lot safer if they were able to do obstacle avoidance and automatic emergency braking, too.”
There are challenges for technology providers, as well, in trying to determine where their technology fits, and how it might be adapted to be useful in the autonomous space.
Simon Blake-Wilson, VP of products and marketing at Rambus’ Cryptography Research division, said as a vendor not steeped in decades of automotive experience, the company sees an acronym soup of different standards bodies trying to work things out. “Clearly you’ll never get anywhere if you try to worry about all of them. Which are the ones we should invest time in? And which areas should we worry about standards versus pioneering? Vehicle-to-vehicle is a good example of where it’s been clear that standards will dominate, with over-the-air updates being much less clear as far as the need for standards. You also have to make judgments as to whether you should ignore vehicle-to-vehicle for the time being because you figure it’s going to take longer, while getting technologies established in the environments that can move quicker. From our perspective, things like over-the-air updates represent low-hanging fruit, whereas vehicle-to-vehicle communications seems like a bigger, long-term opportunity. But how actionable is it in the short term?”
Preparing for autonomous
To be sure all of this requires planning and preparation, and there are a number of key technological shifts that are taking place that represent big opportunities for compelling reasons why engineers should get up to speed on some of these technologies.
“It’s pretty clear that the vision technologies are going to grow in importance just because there is so much interesting data that you can get from imaging,” said Rowen. “Second, I would say that even broader than imaging, these deep learning technologies and what they evolve into are going to be important. We are moving past the phase where you can say, ‘I have developed an algorithm that will recognize vehicles,’ or, ‘I have developed an algorithm that is good at tracking pedestrians.’ People will develop higher level algorithms that are good at creating networks that can be trained to recognize whatever training set you can provide.”
This deep learning, neural network technology is a basic building block not solely associated with vision and probably with pretty long legs. “It will evolve over time, but it is such a powerful general technique that it will become a basic tool in the application builders’ toolkit, just as they think about databases or they think about searching and sorting or they think about tools to manage suffer development software development,” he said. “Part and parcel with that will be increasing attention to data management. How do you get good data sets? How do you train good data sets? How do you select what is the best data set to use in a given training session to develop an algorithm or an application? How do you label data sets, because the quality of your applications now is strictly tied to how large and appropriately labeled your data sets are? That is something a lot of people haven’t really thought about yet very much in these deep learning applications. If I am right and deep learning becomes important, then a lot more people are going to have to get a lot more savvy about the training process.”
Still, autonomous driving technology is one of those few things that will touch probably everyone on the planet, noted Andrew Patterson, director of automotive business development at Mentor Graphics. “It’s not just a technical issue. It’s cultural, it’s legal, it’s software, it’s hardware. Driving and our mobility is at the heart of what we do every day. It’s going to be absolutely huge, and it’s hard to imagine what all the implications and ramifications are at this point. People have classified it as first lose your eyes, then lose your hands, then lose your mind so you’re not actually thinking about it.”
For those of us who have spent more of our living with a driver’s license than without, how long will it take before we can sit and read a book while being driven someplace?
“It’s one thing to imagine taking your hands off the wheel and saying, ‘I’m going to let you steer, but I’m still going to watch the road,'” Patterson said. “The next step is to look the other way and say, ‘I’m not even going to think about what’s going on.’ That’s the ultimate test. When you’re in an airplane, you’re not thinking if the engines are working okay, and if the pilot is doing their job correctly. You absolutely trust that they are, and there’s a big cultural change for many of us before everyone in the car is thinking that same way. That will probably only come over time, when the technology has become so proven and so trusted, as we now do with air travel.”
Designing SoCs For Hybrids
It’s a whole new ballgame when it comes to designing SoCs for hybrid electric vehicles, from regulatory to technical to ecosystem challenges.
Sensors Enable ADAS
Advanced driver assistance requires a careful and complex balance of hardware, software and security.
Tech Talk: ADAS
Why the Advanced Driver Assistance Systems standard is so important and where the potential pitfalls are.
Tech Talk: ISO26262
What can go wrong in designing to this automotive standard.