中文 English

Will Steering Wheels Ever Disappear?

Short answer: Probably not.

popularity

Autonomous vehicles are touted as the way of the future, and the general assumption is ultimately everything will be autonomous. But there are broad segments of the market where autonomy is unlikely to happen.

There is no question that autonomy is possible. In fact, within certain bounds it’s already here. Robo-taxis are available in some cities. Mercedes, Tesla, and Lexus are all touting some autonomous features, and many cars will be capable of autonomy in limited circumstances. But the idea of telling a car where to go, and then sitting back and relaxing, or working while the vehicle takes you there, is nowhere on the horizon.

“There were many assumptions made over the last couple of years based on incomplete data and lots of assumptions about sensors and how expensive they are going to be,” said David Fritz, senior director for autonomous and ADAS at Siemens EDA. This has contributed to unrealistic expectations for autonomy.

Both geography and economics may place limits on where investments will be made towards autonomy. In addition, many autonomy features will apply only to high-end versions. And regardless of location or price point, consumer acceptance may be critical to achieving broad autonomy.

“The market driver is the mass market,” said Gal Carmel, general manager for automotive at proteanTecs. “Otherwise, the OEMs will not make money.”

Personal vs. commercial
The real test of true autonomy would be vehicles that have no interface for driving. That means no steering wheel, no pedals, no gear shift. No consumer cars have yet achieved this level of autonomy. Moreover, it’s not clear they ever will.

When discussing autonomy, there is a big distinction between vehicles for personal use and commercial vehicles like taxis and delivery trucks. The impulse to eliminate labor costs wherever possible means that commercial autonomy has the potential for a much larger return on investment.

This is why there is so much investment in commercial vehicles. There are clear economic savings to be had. “Level 4 is going to come to the trucking industry a lot sooner than passenger vehicles,” predicts Tom Wong, director of marketing, design IP at Cadence.

In addition, the commercial environment can be much more closely controlled. With taxi rides or truck routes, the destination(s) and preferred routing may be decided before the vehicle starts the trip.

Personal vehicles, by contrast, could be more demanding because they need to serve the whims of the driver. On a trip to the grocery store, did you just remember some hardware you need to get on the way back? On a trip to some national park, do you run across an interesting curio shop that you’d like to visit?

Because there really isn’t as much of a question about autonomy with commercial vehicles, most of the following discussion deals with private, personal vehicles — whether owned, rented, or shared — in countries developed enough to afford autonomy.

Can you get there from here?
One of the most obvious challenges to autonomy is geography. If autonomous vehicles could navigate solely by a combination of GPS and inertial sensors, they could go anywhere, as long as there isn’t too long a lapse in the GPS signal.

But the reality of autonomous driving means being able to deal with unexpected events. This goes well beyond just navigation. It includes real-time decision-making, using a combination of cloud-based inferencing and internal technology. Hence, the focus on V2V and V2X communications, where vehicles are expected to communicate with other vehicles and with infrastructure.

With V2V, cars can signal their intentions to other vehicles, sending critical telemetry information to avoid accidents. V2X signals will be able to warn the vehicle of upcoming circumstances, such as traffic incidents or weather changes.

V2V requires technology only on the vehicles themselves, so there is no inherent geographical limit. V2X, in contrast, requires transceivers along the route. Candidates for communications include both IEEE 802.11p and 5G (or other cellular) technology. If existing cellular coverage is any indication of what’s likely to happen, V2X coverage outside major population centers will be sparse or non-existent.

“Ultimately, it comes down to who pays for it and how much you’re going to spend,” said Joe Mallett, senior marketing manager at Synopsys. “You can put the infrastructure in place for everybody, but it’s going to be very costly.”

“In terms of the rural areas, am I saying 100% of the geographic areas will be reached? Not for a very, very long time,” said Vikram Patel, director of product marketing at Infineon, noting it’s not obvious yet how far autonomy will extend. “That’s an ongoing project, where they have these studies to decide how much of the apple they are going to bite into,”

It’s not even certain all of suburbia would be navigable with autonomous vehicles. “Once you get off the freeways, are you going to put infrastructure all the way to every cul de sac?” asked Mallett.

Countries such as China, which are just building up their highway infrastructure, are adding lanes specifically for autonomous vehicles. This will allow for the installation of any needed infrastructure from the ground up, which is much easier than trying to upgrade millions of miles of existing roads.

“Our customers are designing vehicles and systems to go on these expressways,” said Kurt Shuler, vice president of marketing at Arteris IP. “They have dedicated lanes between big cities. But I don’t think in the U.S. we’d be doing dedicated lanes.”

Others agree. “They’re building all these new cities where you can create infrastructure from scratch, as opposed to trying to do it in downtown Boston, where you have established infrastructure that you would have to update,” observed Frank Schirrmeister, senior group director for solutions and ecosystem at Cadence.

Maybe rural isn’t so hard
This is how the narrative has gone for a while. In many rural locations, a fully autonomous car with no driver interface would be expected to pull over to the side of the road and stop if further progress is deemed unsafe, returning the car to manual control.

However, this may not be the case. All in all, autonomy in rural areas is easier than in cities because there are far fewer interactions and corner cases to contend with.

“The most challenging environment is going to be the urban environment,” said Willard Tu, senior director, automotive at Xilinx. “There could be a lot higher traffic density and more objects to track, making it more difficult for any onboard computer or cloud server to help manage.”

In rural areas, there are typically far fewer cars and pedestrians. Some of the use cases will be different, like navigating roads that aren’t straight, that have no center line or clear edge lines, and that might not even be paved — making it necessary to dodge potholes or navigate washboard surfaces (although some urban areas, like Portland, Oregon, still have some unpaved roads).

As it turns out, for the better rural roads, some are finding that their cars already can handle the drive autonomously, given favorable conditions. “I live in an extremely rural area in the mountains, and, as long as the white and yellow lines are clearly marked, I could drive my entire road without touching the steering wheel,” said Michele Kinman, founder and director of public relations at The Edge Marketing.

“We can detect the edges of the road or an animal running in front of you,” said Fritz. “And we’re doing that right now using 20 to 25 watts,” as contrasted with 4 kW experienced by some trying to achieve full autonomy.

The installation of infrastructure to support V2X is lagging behind the progress to autonomy, so developers are attempting to make autonomous vehicles as self-sufficient as possible. V2X may help in the future, but it may not be required in the short-term. So rural driving may achieve autonomy before urban driving, assuming that OEMs prioritize investing in the use cases that apply to rural roads only.

“Still, the priority may be a lot lower in that area,” cautioned Tu.

Even in cities, some level of autonomy may be possible. “You can operate safely in urban areas with or without smart-city V2X infrastructure,” said Fritz. “You can operate in rural areas without any smart-city infrastructure or V2V of any sort.”

Tu concurred. “The vehicle has to be able to do everything by itself,” he said. “If the V2X connection goes away, it should be able to do a lot of the functionality — maybe not as robustly, but it should be able to manage.”

Who’s going to pay for that?
Another limiter for autonomy is economic. Self-driving or driver-assist features garner a lot of headlines, but they can be expensive.

“There are a lot of vehicles that people just want for transportation,” said Tu. “They’re going for $15,000 to $20,000. If you were to add this autonomous feature, that might be half the vehicle price itself.”

Even if autonomy can be economically implemented, there is an enormous level of technology investment that must be repaid before the automotive OEMs can start making money.

Some suggest that there’s something of an ego-based race going on at present, which can’t really last. “There’s this prestigious race between the OEMs,” observed Robert Schweiger, director of automotive solutions at Cadence. “Who will be the company that first achieves Level 4 or 5 autonomy? We all know that the investment that goes into that doesn’t get a real financial benefit back.”

When cooler heads prevail, high-end features tend to go where customers are willing to pay a premium for them. And that means high-end cars.

This also helps to support a pricing tier that can support a premium at the high level because there are lower-priced tiers below that. “If everybody’s an L5, then what’s your differentiating trim level?” asked Mallett. “Maybe you have a fold-down bed. Maybe you have a TV system.”

This wouldn’t be the first time some vehicles would have more bells and whistles than others. Once upon a time, features like power windows were considered high-end options. Today, such features have moved down into the affordable range — and yet it’s still possible to purchase cars without them.

Fig. 2: Autonomy offers the promise of doing things other than driving while in transit. Source: Siemens

Fig. 1: Autonomy offers the promise of doing things other than driving while in transit. Source: Siemens

Similarly, it is expected that various ADAS capabilities will start out on high-end vehicles, and then gradually over time migrate down to more moderately priced cars. This allows those early movers, who will pay higher prices, to help pay off the investment needed to create the technology.

“Once they can amortize their R&D, then it goes into mass production because the patents have probably all expired, and now it’s making money off the revenue,” said Amol Borkar, director of product management and marketing for Tensilica vision and AI DSPs at Cadence.

Full autonomy sets a high minimum bar for features. The various in-cabin goodies still could be part of the price rationalization because they’re for convenience and comfort. But an autonomous vehicle must be safe, and so anything safety-critical would have to have a minimum level of capability in order to exist at all.

“If you look at some of the current tier levels, you see that safety is becoming more ubiquitous across all vehicles,” noted Mallett.

A common subset of features may pervade even the bottom tiers. “Things like a backup camera and checking for obstacles and pedestrians on the backup camera are becoming standard,” said Borkar.

Two possible scenarios
Will the price of full autonomy come down enough over time to where the lowest-priced vehicles can have autonomy without it being sold at a loss?

There are two theories on how this will play out, and they depend on perceptions about whether the hardware needed for autonomy is too expensive.

One of those theories involves the notion of the software-defined vehicle. Such a car has the hardware necessary for many features, but those features need to be enabled by software. This is possible because of the falling price of electronics. “Hardware technology nowadays is not expensive,” said Prakash Madhvapathy, product marketing director, Tensilica audio/voice DSPs at Cadence.

“The delta in autonomy, from a hardware and system point of view between premium cars and the lower segments, is not that big,” said Carmel.

The idea is that the ability to up-sell or “rent” features after the vehicle has been purchased can serve as a new source of revenues. This appears to be one of the main drivers of the huge investment in automotive technology.

“Everybody’s working toward what we call the software monetization concept,” said Tu. “By 2030, you’re going to see vehicles have hardware that the consumer may not be using, but that they’re going to be able to use it if they want to pay more.”

If this happens, then the lowest-end cars might indeed become loss leaders for OEMs on the hopes that drivers will upgrade either the driving capabilities (like levels of autonomy) or the comfort and convenience features.

A counter-theory suggests that, as inexpensive as hardware might be, it’s not free. “What OEMs are most concerned about is that, a couple of years ago, there was a McKinsey report that essentially said the further we get into autonomy, the higher the percentage of profits in the vehicle is going to go to the chip companies and not the OEMs,” said Fritz. To recapture some of that, he noted, “Companies are already announcing that they’re looking at building their own silicon.”

So this model holds that hardware will be modular, and located somewhere easy to access, such as between the front seats. Some number of boards could be installed, depending on the level of sophistication desired. An entry-level car may have one simple board, while a fancier version might have multiple boards.

It might be possible to upgrade the hardware, as well. “If you want a premium vehicle that could do it all, you can add another two or three boards,” said Fritz.

Low-end vehicles, then, would have the minimum level of ADAS or other safety hardware required by government mandates. That would keep prices within range without having to make them loss leaders. “Maybe there’s not a backup camera, but there are definitely sensors that let you know somebody’s in your blind spot,” said Mallett.

Regulatory mandates will impact some feature choices. “Government definitely has a role,” said Shuler. “There have been discussions around how much the industry should be able to find their own rules versus how much the government, at the federal level or individual state levels, should regulate. Industry would like to create their own rules and have those apply to all 50 states.”

“Safety is what government can control, and they will probably mandate certain things or incentivize the automakers to implement safety features,” said Ted Chua, product marketing director in the Tensilica IP group at Cadence.

Safety-ratings agencies could have a similar impact. “What’s also happening is the safety-ratings companies are encouraging automakers to do certain things,” Chua added.

And they’re not the only ones with influence. “The one thing not to ignore is what the insurance companies think,” observed Madhvapathy.

Required capabilities don’t have to be fancy, of course. “Because it’s a mandate, I’ve got to have it be good enough. I don’t have to be the best in class,” said Tu. “The premium brands will have the best in class.”

Upgrades in both hardware and software still could provide additional revenue opportunities after the initial purchase. Full autonomy therefore would not be enabled on those low-end vehicles without further hardware upgrades.

It all comes down to when the vehicle has to cry for help from a driver. Given a tricky situation, “The systems themselves will say, ‘I can’t handle this,’ and give you back control,” said Fritz. “What the high-end vehicles can handle before they throw up their hands will be far more.”

Gaining public trust
The final condition that must be met for autonomy to be a commercial success is for consumers to accept it. “How many people are really eager to have an autonomous car?” questioned Michael Frank, fellow and chief architect at Arteris IP.

There are a number of cultural and societal considerations that determine acceptance. “There will be pushback on autonomy, just like in any new technology,” predicts Tu.

Schirrmeister referred to a study Cadence did regarding consumer preferences. “People are more negative on autonomy in rural areas than in the city,” he said. “In the case of China, sentiment is a net positive, whereas in the case of Germany, it’s the most negative.”

Competition forces OEMs to pay attention to consumer demands. “In this case, there are going to be enough competitors,” said Shuler. “I don’t think it’s going to be like Facebook or Google where, if you want to search or do ads on the internet, you deal with us.”

To the extent that monotonous commute driving can be left to the vehicle as an alternative to public transit, there’s likely to be some broad level of acceptance of autonomy. “If you look at the U.S., at least in our area (Silicon Valley), they don’t have anything that could really be called public transport,” said Frank. “But if I go to Europe, no matter which major city it is in any country, I don’t really need a car.”

Beyond commuting lies the realm of personal preference, and that involves a broad spectrum of opinions. Some people enjoy driving immensely, and commuting aside, they like to be able to get out and clear some carbon out of their heads and hearts. There are also people who very much dislike having to drive. So there’s no one rule for everyone.

But when the momentum is overwhelmingly toward autonomy, will the desires of driving aficionados matter? “I still think the vast majority of humans are going to want control as long as they can afford control, and the OEMs are sensitive to that,” said Fritz.

There’s also the issue of privacy and the possibility that cars will be yet another way to harvest data. If you’re working while the car drives itself, will in-cabin cameras potentially be sending out confidential work information?

Feelings about privacy vary widely. Many people claim not to like being surveilled, and yet given the option of something free, they’ll grumble and give away their data — at least in the United States. Europe is more concerned about privacy, while China is less concerned.

“People like their free stuff,” observed Shuler. “They like convenience, and they’re willing to give up so much information — at least in this country.”

There are also more granular geographic distinctions. Urban driving is much less interesting — and cities are where public transit is more likely to be an option (especially outside the U.S.)

In rural areas, by contrast, manual driving is likely to remain a preference for much longer. “Every teenager in a rural area probably looks at a car as freedom,” said Tu. “They don’t have mass transit, so, for them a car is about speed, it’s about looks — all those things that we look for when buying a car today.”

The biggest hurdle for consumers will be whether they perceive autonomous vehicles as safe. And that might mean OEMs need to under-promise and over-deliver. “The industry has to has to be credible and trustworthy,” said Shuler. “If you hype it up too much and people start thinking autonomy is possible today — and then they realize it’s not — then it becomes disillusioning.”

Conclusion
You’ll hear a variety of opinions on if and when “Minority Report” types of vehicles will appear. But there’s still lots that the automotive industry will need to learn about what works and doesn’t work.

And humans will adapt, too. “Huge societal changes are going to happen because of this,” predicts Shuler.

As to that steering wheel? “Over a long period of time, that steering wheel will fade,” said Tu. “It might be something that retracts into the vehicle, there just for emergencies.”

But it also may never completely disappear.



3 comments

Reynold Funk says:

How long before software-enabled features are hacked, either for own use/black market sale, or for nefarious purposes? Present examples of capabilities are extremely concerning.

s nedunuri says:

The article seems to suggest that having the steering wheel around for situations that autonomy can’t handle. This is fine if the vehicle is in some low traffic street where there’s opportunity to pull aside and stop. But what if its in a high speed situation like on a freeway, or on a major road with no roadside stopping or parking? And in those situations, can a person really just “take over”? If they’re in the middle of watching a movie, or arguing on the phone, or just plain napping, how much time will it take a person to register and comprehend the situation and then figure out the best course of action? Think of the Air France flight in 2015 when the auto controls failed, the pilots were unprepared to take over.

Another point that the article partially acknowledges. Full automation isn’t simply a matter of piling more processors and sensors onto a vehicle (even assuming those sensors were perfect, which they aren’t). It requires that the autonomous system be able to interpret the scene its looking at and apply some rudimentary reasoning to the situation to determine the best course of action. No existing autonomous system today can do that. All the classic examples come into play here – people obscured by other vehicles or objects, what to do if the vehicle is at a red light but another vehicle is about to collide with it, how to handle situations in which the road signs have been removed (or worse yet the incorrect ones left in place) due to construction, etc. In all these situations human drivers can interpret the scene and carry out an appropriate action deciding which rule to bend or break or doesn’t apply depending on the situation.

paul adriaan kleimeer says:

I thought ai was going to eventually encounter every possible situation and deal with it. Even human reaction time can’t deal with things it can’t see or predict.Crash avoidance needs an time distance lag to react like crash avoidance in airplanes . Lidar cant stop you running down dog when it runs out from infront of a stationary object and the brake reaction is not fast enough to stop a collision. My real concern is when all accidents are avoided like there are no more incidents and 100% safety will insurance still be required or can we assume it will no longer be require.

Leave a Reply


(Note: This name will be displayed publicly)