VW to eliminate worst road hazard: drivers
In 2028, the car you won't drive won't be yours
Hot Chips Soon you won't own a car, but one will come to you on its own when you call it, then whisk you away in perfect safety without you having to drive it — and that day may be closer than you think.
"If you ask, 'Is it a future story you're telling us?' No, it's not," said Burkhard Huhnke, executive director of the Volkswagen Group of America's Electronics Research Lab in Palo Alto, California, speaking to the Hot Chips conference earlier this week about what he calls "autonomous cars".
Although fully autonomous cars won't appear for about 20 years, Huhnke says that his research group is well on its way. "We are looking into some of the applications coming pretty soon. Traffic-jam assistance, for instance, automatic parking — we have park-assistance already introduced — collision-avoidance systems, and an emergency braking system that brakes automatically if it recognizes an obstacle that's in a specific speed range."
A "speed range" that includes, one assumes, not moving.
Huhnke cited two main motivations for autonomous cars: safety, and the elimination of what he identified as "annoying" driving — meaning, for example, traffic jams and long, boring stretches of open road.
He noted that current automotive-fatality figures in Germany hover at around 5,000 per year. "Over ten years you're counting pretty soon 50,000 people — and that's something that's really crazy. I think someone compared that with airplane crashes — would we accept that two airplanes would crash per day over a year? No, we would not accept that. But obviously we accept that in our daily traffic experience."
The reason for the vast majority of traffic accidents, he said, is human error. Of all the many and varied reasons for crashes, he cited studies which have shown that "84 per cent are misjudgment by the driver."
The solution is obvious: get rid of that error-prone driver.
"Here at Hot Chips you create consistently great chips and computers and you would not accept that the user would create so many problems. But as we all know, the user is usually the reason for the problems with the computers — and that's also the same with the cars."
He noted, though, that per-kilometer traffic fatalities have dropped precipitously in Europe in recent years, thanks to passive safety features such as seat belts, crush zones, and protected passenger compartments. Helping that decline have also been electronic safety features such as anti-lock braking system, electronic stability controls, and traction-control systems.
Those electronic systems have also added to cars' chip counts — according to Huhnke, there are "almost 50" CPUs in an average Volkswagen today.
But more electronic safety features are needed, he said. Active safety features such as adaptive cruise control, lane-departure warnings, and blind-spot detection, when taken together, could result in a 20 per cent reduction in fatal crashes, he estimates.
But to become truly autonomous, "The car needs to learn to see, [have] sufficient technology to understand its environment, to make the right decisions." Such a car would also be able to communicate with other cars, using what Huhnke called "wireless 'all-way' connection in a smart way."
As an example of the advantages of car-to-car communication, Huhnke envisioned a blind curve, around which a car has broken down in your lane. Rather than merely speeding around the curve and plowing into the rear of that stopped vehicle, "the cars can communicate and the driver can be warned."
Wireless-equipped cars could also communicate with traffic signs, which could provide information of road and traffic conditions — and if your car were fully autonomous, it could regulate its driving without you even knowing that it had received any external help.
Huhnke said that his group wanted to find out if
drivers passengers in autonomous cars would feel safe: "If you have an autonomous car driving ... do you trust your car? Do you really press the autopilot button and let the car drive you at 60 miles per hour?" So they conducted a study — and were surprised by the results.
"We created a car with a second steering wheel in the rear where the driver couldn't see it," he told his audience. "He or she pressed the autopilot button and thought the machine would really drive without human help. Someone drove in the rear seat without being recognized by her or him. Well, you couldn't imagine: after a few seconds, they already took the newspaper and read the news articles. So they trusted already the machine, which was great."
Huhnke's group then pushed its luck: "We also initialized some emergency situations: 'So please, go back to your steering wheel and take over, we need some help from you,' and they did it. They put the newspaper back, and just controlled the car through the situation. Then what did they do? Immediately press the button and start it again — it was really amazing."
Autopilots and angels
Huhnke defined two different types of assistance for two different scenarios. "Under-challenging the driver — it's really boring, you're just in a traffic jam — and over-challenging the driver, when it's extremely dangerous." For those under-challenging times, the autonomous car goes on autopilot. For the over-challenging time, it becomes, in Huhnke's words, "a safety angel."
To help in both types of situations, a car needs sensors. Huhnke identified a few that already exist: multi-beam lasers, rear and front cameras, radar, and "supersonic" sensors. Add to those car-to-car and car-to-data-repository wireless connections, and the holy grail of autonomy becomes attainable.
Huhnke gave one San Francisco–related example of autonomous-car convenience. "Imagine you're at the Fisherman's Wharf parking garage. I usually have the kids with me and they're really annoyed by waiting. And we need to drive up to the fifth floor [where] finally we can find a parking spot. It takes you at least 25 minutes, at least," he said.
"But wouldn't it be great if you just press the button of your iPhone application telling your car to find you a parking spot?" If your car is not so equipped and Huhnke's is, "You still have to drive up the parking garage. I go shop."
You also wouldn't have to know where you parked your car in the autonomous future. First, because it could take care of that on its own, and second, because, well, it wouldn't be your car — it'd be part of a fleet of vehicles that you could summon whenever you need one. To illustrate this concept, Huhnke showed a cutesy 2008 video created by Volkswagen that describes its vision of what car usage will be like in 2028:
The Volkswagen Ego in that video, by the way, was summoned from its charging station, where it had zipped off to recharge itself when it was feeling low.
As rosy as the picture Huhnke painted might be, he did note that there are some obstacles that the inexorable march of technology can't by itself overcome. "We'd like to find smart solutions," he said, "but the customer, of course, needs to be willing to pay for those solutions at the end of the day."
There's also the simple fact that cars, once purchased, hang around for quite a few years. "You might ask: 'Okay, if this technology is going to be introduced into the market, how many cars do we need, what's the market penetration for that to increase the safety?'"
The answer to that question depends on many factors — not the least of which being the health of the economy, the prices of autonomous cars, and public acceptance. Even if such cars became immediately available across all automotive sectors, Huhnkee said, "You need to wait a few years, you need to have lots of cars delivered with these kinds of technologies. It takes some time."
And, of course, if autonomous car-to-car communication is to be at all viable, "You have to have a very secure network for the safety features." ®
The cars in Huhnke's autonomous future will, of course, be electric — but he pointed out one law of thermodynamics that's going to be mighty hard to overcome: when fueling your car with petrol, you can pump 1,000 kilometers of energy into it in one minute. A plug-in electric car? Not so good: one kilometer per minute.