Cars That Think iconCars That Think

Delphi and Mobileye say they'll introduce a Level 4/5 self-driving car system before the end of the decade.

Delphi and Mobileye Will Offer a Fully Self-Driving System by 2019

Delphi, an American auto supplier, and Mobileye, an Israeli self-driving firm, say that by 2019, they expect to introduce a fully self-driving system—one capable of taking the wheel and holding on to it for the duration of most trips. 

The system will be what robocar wonks call a “Level 4/5” system. Level 4 means the car can drive itself from start to finish except under unfavorable circumstances. Level 5 is the same, but without the caveat. That’s a very tall order, even for three years from now. Or four, if you count the time it’ll take automakers to integrate the system into their vehicles.

Not even Google, which jump-started the robocar trend, now predicts that throw-away-the-steering-wheel capability will come anytime soon. Instead it’s talking five years for sunny climes and exquisitely mapped roads, and up to 30 years for snowy backways in the Yukon.

Even so, in today’s market, Delphi and Mobileye are not so much leading the pack as scrambling to retain their positions. In the past year, a number of automakers have taken steps to bring the technological guts of robocars in house, and that threatens to leave suppliers out in the cold.

Just last week, Uber bought the robotruck startup Otto and said it would work with Volvo to start a commercial, human-monitored ride-hailing service in Pittsburgh this summer. The week before, Ford Motor Company and Baidu, the Chinese search company, invested in Velodyne, a leader in LIDAR—a key, laser-based sensor technology. In July, Tesla Motors severed its relationship with Mobileye, ostensibly to take more control over its own robotech projects, though there were also other reasons for the split-up. And back in March, General Motors acquired Cruise Automation

At a conference call today and in comments beforehand, Delphi and Mobileye said they would serve smaller auto companies, which might lack the resources to develop a robocar outfit or buy one ready-made. They said they planned to invest “hundreds of millions” of dollars in the venture.

In the meantime, to keep enthusiasts stoked, the companies will show off the current version of their system in January, at CES, the expo formerly known as the Consumer Electronics Show.

Uber Will Start Driverless Service in Pittsburgh—This Month

Driverless taxis are coming—right now. Later this month Uber will offer the world’s first ride-hailing service in Pittsburgh, using a test fleet of 100 Volvo XC90 SUVs.

I admit that I didn’t see it coming this fast. Only yesterday I wrote about a new trial of self-driving minibuses in Helsinki, and it seemed pretty darn forward looking at the time. But those vehicles are really slow, they ply the same little route repeatedly, and they stop at every stop.

That was so yesterday. Today Uber is talking about driving to points specified by whoever jumps into the rear seat. And it’s no mere one-off deal with Volvo, either: The two companies will collaborate on technology, including mapping, and Volvo intends to put what it learns into its other cars.

To be sure, two other outfits are already experimenting with robotaxis in Singapore—NuTonomy and Delphi. But the range of activity there is, for now, rather more limited than Uber’s planning in Pittsburgh, and far fewer cars are involved.

Uber also has bought Otto, a year-old startup filled with émigrés from Google and other robocar powerhouses. The idea is to equip Uber to deliver cargo within cities.

Most companies that are interested in robotic driving are now saying that they, too, hope to use it in car-hailing services. Ford—which earlier this month said it would bring true driverless cars to market in 2021—has also singled out ride sharing; it, too, is working with Uber.  

One beauty of ride sharing is that it lets Uber emulate Google by testing things out with a professional driver at the wheel, ready to take over if need be. But here the driver is doing two jobs—backing up the automation and conditioning customers to robodriving. A tablet computer in the back seat will explain what’s happening to the customers.

“The goal is to wean us off of having drivers in the car, so we don’t want the public talking to our safety drivers,” Raffi Krikorian, Uber’s engineering director, told Bloomberg

Ford autonomous car with Velodyne lidar sensors

Ford and Baidu Invest $150 Million in Velodyne for Affordable Lidar for Self-Driving Cars

Elon Musk may not think that lidar is a necessary sensor for self-driving cars, but pretty much everyone else in the inudstry disagrees. With its long range, impressive accuracy, wide field of view, and near-immunity to the vagries of ambient light and weather, lidar offers a volume of high quality data that's hard to get with cameras or radar. Companies like Ford and Baidu see lidar as an integral sensor on their near-future autonomous car projects, and to support that vision, they've just invested US $150 million in Velodyne, the company that makes the best lidar sensors on the planet.

Ford and Baidu are both very specific about what this $150 million investment is going to do for them. Ford wants Velodyne to "quickly mass-produce a more affordable automotive LiDAR sensor" so the company can launch a fleet of self-driving ride-sharing cars by 2021. Meanwhile, Baidu is hoping to "advance Velodyne’s development of increasingly sophisticated LiDAR sensors."

Read More
An EasyMile EZ10 self-driving bus in Helsinki.

Helsinki Tries Self-Driving Buses in Real Traffic

Two small, French-made EasyMile EZ10 buses are now driving themselves in real vehicular traffic in Helsinki. A human being is around, just in case

The month long trial, in the south of the city, follows an earlier experiment in a neighboring municipality on streets limited to pedestrian traffic, reports YLE, a local news service. Other places have tried this hothouse-flower experiment, notably in the Netherlands, Switzerland, China, and Britain. Next month the service will move to a municipality immediately east of Helsinki, and then on to Tempere, which is 160 kilometers (99 miles) to the north.

Read More

Millions of Volkswagens Can Be Unlocked by Hackers

Millions of Volkswagen cars could in principle be unlocked from a distance by hackers, according to a new report. It’s one more strike against a German company that’s had more than its share of bad news, what with VW’s admitted cheating on diesel-emissions tests.

But there’s plenty of bad news to go around: The report notes that many models from other companies have vulnerabilities of their own. And the problem is worse with older cars, designed before carmakers gave much thought to cybersecurity.

Flavio D. Garci, David Oswald, and Pierre Pavlidès at Britain’s University of Birmingham wrote the report, together with Timo Kasper of Kasper & Oswald, a German security firm. They released it [PDF] last week at the 25th USENIX Security Symposium, in Austin, Texas.

Kasper told the BBC that the researchers informed Volkswagen of the vulnerability in late 2015 and that the company was taking steps to mitigate the problem. But he said there are “at least 10 more, very widespread” vulnerabilities affecting other car brands, which the researchers won’t publish until the manufacturers have had time to do something about them.

The report found two weaknesses. The simpler hack involves discovering which of VW’s small handful of “master key” codes is used by a particular remote key fob by eavesdropping on a single signal from that fob to its car. That code, together with the algorithm, can then be used to unlock that car. A lot of cars are involved: The VW Group sold about 100 million cars from 2002 until 2015, and the researchers estimate that “the vast majority is vulnerable to the attacks presented” in their study.

The more complex hack allows the bad guys to take over cars using the Hitag2 rolling code method, used in cars made by Chevrolet, Renault, and Ford,
among others. Rolling code changes with each use, but the algorithm that changes it can be inferred “with four to eight rolling codes and a few minutes of computation on a laptop,” the researchers write.

The paper shows a picture of the researchers’ handmade RF transceiver, which is basically an Arduino processor and a few other parts, all fitting on a circuit board hardly larger than the battery to which it is attached. “Our simple setup... costs ≈ $40,” the paper continues, “is battery-powered, can eavesdrop and record rolling codes, emulate a key, and perform reactive jamming.”

The main challenge to the would-be hacker is in getting a number of distinct rolling codes, though there’s no need that they all be consecutive codes. But the researchers helpfully note that it should be easy to induce the naive car owner into giving up that information by punching his remote fob again and again.

“The adversary could selectively jam the signal during the final checksum byte (which is predictable),” they write, referring to the data at the end of a transmission. “In this case, the vehicle ignores the rolling code, but the adversary nevertheless obtains the keystream. The victim would hence notice that the vehicle does not respond, and instinctively press the button repeatedly. After having received the fourth signal, the adversary stops jamming and the remote control operates normally from the victim’s point of view.”

How might bad guys use this trick to do evil? The authors note that they could take something out of the car or put something (or someone) into it, perhaps giving the owner a nasty surprise later, when he’s all alone on the open road. And of course once inside a parked car the hacker would have the safety and leisure to continue the break-in with tools applied directly to the electronic systems under the dashboard.

The remote fob has been a security bugbear for years, though the focus of worry keeps on changing. In 2014 IEEE Spectrum reported that thieves were making clones of fobs they’d “borrowed” while a car was in the shop, being repaired. But that takes planning and, perhaps, accomplices. If you can get what you want with a homemade circuit board, you are better able to keep the secret—and all the loot.

Today's Electric Car Batteries Can Get You There and Back Again (Usually)

A nagging problem with electric vehicles is range anxiety: the creeping suspicion that you might not make it to the garage before your batteries die.

Now a study published in Nature Energy concludes that most people rarely go farther than a single charge allows and that a significant minority practically never do. That opens up a big segment of the car market—and that’s just with today’s batteries. Tomorrow’s will be far more capacious.

Jessika E. Trancik and colleagues at the Santa Fe institute and MIT modeled driving behavior in a new way, by collating databases. From the National Household Travel Survey they obtained information on a large number of trips by all sorts of cars, not just EVs. To characterize the second-by-second velocity of each kind of trip, they referred to a set of GPS-based measurements. Finally, to account for energy consumed in heating and cooling, they factored in a historical database of local weather conditions. This gave them a range of probabilities for the distance traveled and the battery capacity used.

Read More

Retracted: GM-Lyft Self-Driving Taxis Could Launch in 2019

Editor’s note: Our story was based on the premise that the advanced radars discussed below were intended for autonmous cars. GM contacted IEEE Spectrum after publication to say that the radars are not intended for autonomous vehicles. According to GM:

The FCC filings referenced in the IEEE Spectrum story are not part of our autonomous vehicle development program.  They are related to further advancement of technologies featured on our vehicles today. 

The original story continues below:

Read More
Innoviz Technologies

Israeli Startup Innoviz Promises $100 Solid-State Automotive Lidar by 2018

Yesterday, Israel-based startup Innoviz Technologies announced a $9 million Series A round to fund the development of a high performance, low cost, solid-state lidar to enable fully autonomous driving. Lidar is recognized by most (but not all) groups developing autonomous vehicles as a necessary sensor, because it provides high resolution and long range 3-D data that are independent of lighting conditions and most weather. However, the cost and complexity of currently available lidar systems make them impractical for most consumer vehicles, which is why there's been so much interest recently in developing an inexpensive and reliable (but technically challenging) solid state lidar.

Innoviz is promising to deliver, within the next year or two, a “game changing high definition solid state lidar (HD-SSL) based on breakthrough technology which offers superior performance and accuracy.” The device is expected to have a wider field of view, and higher resolution in both axis and long range sensing than today’s lidar systems. But Innoviz says it will be significantly smaller, both in size and price. 

Read More

Chris Urmson Leaves Google Car Project

Chris Urmson, head of Google's self-driving car project, has announced his departure without saying what he'll do next. 

If I can find another project that turns into an obsession and becomes something more, I will consider myself twice lucky,” he said in a final statement.

Read More
The early days of autonomous vehicle research

The Electronic Highway: How 1960s Visionaries Presaged Today's Autonomous Vehicles

In July 1969, IEEE Spectrum published an article called The Electronic Highway, by Robert E. Fenton and Karl W. Olson, two engineers at Ohio State University who were working on ways to make vehicles operate autonomously when traveling on major highways. Nearly 50 years have passed, which is practically forever in a technological context, but what's striking about this article is how many contextual similarities there are between the past and the present.

(For more about the history of intelligent transport, make sure to read our feature on Charles Adler, who was working on intelligent traffic control systems in the 1920s.)

The specific solutions that Fenton and Olson propose are a bit outdated, of course, but the problems that they discuss and the future that they look forward to have a lot in common with those peppering current discussions on vehicle autonomy. IEEE members can read the entire article here. We'll take a look at some excerpts from it, and talk about what's changed over the last half century, and what hasn't.

Read More

Cars That Think

IEEE Spectrum’s blog about the sensors, software, and systems that are making cars smarter, more entertaining, and ultimately, autonomous.
Contact us:

Senior Editor
Philip E. Ross
New York City
Assistant Editor
Willie D. Jones
New York City
Senior Writer
Evan Ackerman
Berkeley, Calif.
Lucas Laursen

Newsletter Sign Up

Sign up for the Cars That Think newsletter and get biweekly updates, all delivered directly to your inbox.

Load More