Cars That Think iconCars That Think

Tesla Reveals Its Crowdsourced Autopilot Data

Tesla has unveiled some results of its crowdsourcing efforts among users of its Autopilot feature, introduced just six months ago, and the data give some hints on how the feature performs.

For instance, Autopilot centers the car in its lane with the obsession of a chronic picture-straightener:

Read More

Autonomous Vehicles Traveling in Convoys Will Run Into This Inevitable Tradeoff

Autonomous driving experts practically salivate over the vision of cars forming neat little convoys that travel at a constant speed down a highway. And for good reason: These synchronized fleets could reduce traffic fatalities, free up room on the road for additional vehicles, and help everyone get to their destinations faster.

The problem is that as a convoy grows in size, it invariably runs up against limits that hurt performance. Most notably, the number of messages that vehicles can successfully exchange tends to drop as more cars join. Every additional car requires the generation of more messages amongst the group until eventually, some of this digital traffic starts to collide.

Ignacio Llatser, a researcher at Technische Universität Dresden in Germany, is trying to find the sweet spot at which multiple cars traveling together can share key information without drowning each other out. In particular, he has conducted a series of simulated road tests to optimize the way vehicles exchange data when traveling in a convoy. He outlined his work on Tuesday at the IEEE International Conference on Communications in Kuala Lumpur, Malaysia.

First, his group set out to characterize the fundamental issue: that vehicle-to-vehicle messages are less likely to arrive at their intended destination as a convoy grows in size. To do this, the team used a vehicular simulator to test how successfully messages were transferred between cars as they traveled in convoys ranging from six to as many as 40 vehicles. They ran 10 simulations of 30 seconds each and measured how reliably messages were delivered in each configuration. 

After the tests, the TU Dresden group found that a measure of reliability called the node coverage ratio dropped from nearly 100 percent when there were just six cars in the convoy to just below 85 percent in some cases when 40 cars traveled together.

Next, the researchers found that spacing messages out (in other words, lowering the frequency with which they are broadcast) in larger convoys could prevent this dip in reliability to some extent. In 40-car convoys, they discovered, it’s possible to keep the node coverage ratio from falling below 85 percent by broadcasting no more than one message every 200 milliseconds.

However, sending messages less frequently causes another problem. It increases data age, or the interval between the current time and the instant when the last received message was generated. As data age increases, it increases the likelihood that vehicles traveling at high speeds won’t receive critical information in time to make adjustments. Developers will need to carefully consider this tradeoff when designing autonomous cars for real life scenarios.

It’s still not clear to Llatser and his colleagues how exactly a 15 percent drop in reliability would affect autonomous driving in real life. Would it cause accidents or have no impact whatsoever? And they haven’t, to this point, worked out what are safe data ages for cars traveling at various speeds. But they are studying that question now.

The problem is an important one because convoys are quickly becoming the hot new research area for autonomous driving experts. In the past, many researchers have focused on a related issue: platooning. A platoon is a group of autonomous vehicles arranged in a single lane that follow a lead car; the lead car makes executive decisions about speed and direction.

Companies have already begun to discuss standards for how vehicles in a platoon will communicate with one another. In April, six platoons of self-driving trucks made headlines when they successfully completed road trips to Rotterdam, the Netherlands, from various points in Europe.

Unlike platoons, convoys operate across multiple lanes with no designated leader. Each car makes its own decisions. Llatser’s work represents one of the first close looks at the tradeoff in how messages are exchanged between vehicles when they are traveling together but aren’t arranged neatly, one behind the other.

The TU Dresden results are particularly timely because manfacturers have yet to agree on standards for convoys. This leaves open the possibility of adjusting message frequency in order to boost reliability or lower data age to suit different road scenarios.

But before they decide anything, experts will need more information. 

Later this year, Llatser will participate in a project called AutoNet2030, where researchers will conduct the first demonstration project of a convoy on a closed circuit. That demo, which will take place in Sweden, will feature two cars and two trucks traveling together and executing maneuvers such as merging.

Otto Self-Driving Truck Company Wants to Replace Teamsters

Ottomotto, a newly unveiled startup formed by veterans of the Google car project, is planning to provide self-driving technology to today’s long-haul trucks.

It’s a logical first application. Semitrailers spend most of their time on highways, and highway driving is by far the easiest sort of driving to automate. So close is the industry to that goal that companies like Tesla and Daimler already feel the need to use buzzers and other tricks to prod drivers away from their daydreams when their hands are off the steering wheel for more than a few seconds.

Anthony Levandowski left the Google car project to start the company back in January, the Wall Street Journal reports. Lior Ron, who had been in charge of Google Maps, also joined, along with 39 others, including some from Tesla and Apple—both of which are also working on robocar technology.

Levandowski told the Journal that his company, which he calls Otto for short, would seek a competitive advantage by retrofitting existing trucks rather than putting self-driving systems into new ones. Otto, so far mostly a self-funded operation, is working with three Volvo trucks. 

Read More

Autonomous Mini Rally Car Teaches Itself to Powerslide

Most autonomous vehicle control software is deliberately designed for well-constrained driving that's nice, calm, and under control. Not only is this a little bit boring, it's also potentially less safe: If your car autonomous vehicle has no experience driving aggressively, it won't know how to manage itself if something goes wrong. 

At Georgia Tech, researchers are developing control algorithms that allow small-scale autonomous cars to power around dirt tracks at ludicrous speeds. They presented some this week at the 2016 IEEE International Conference on Robotics and Automation in Stockholm, Sweden. Using real-time onboard sensing and processing, the little cars maximize their speed while keeping themselves stable and under control. Mostly.

Read More

Bosch Takes Me For A Ride, On An Electric-Assisted Bike

Riding an electric-assisted bicycle gives me a whole new perspective on Manhattan. For one thing, IEEE Spectrum’s office on Park Avenue suddenly seems awfully close to the East River. I hadn’t visited it in a decade.

For another, it makes me feel like Superman.  My Scott-made, Bosch-electrified Scott Sub Evo 10 bike does nothing on its own, it only amplifies my own effort, and when I put it on “Turbo,” the highest level, I shoot forward like Lance Armstrong on steroids. As it were.

And by they way, electric-assist bike technology can now be shrunk so deeply into the frame of a bike that it can perform “motor doping” that’s detectable only with special thermal cameras. Racing results may already have been perverted.

What’s the demographic for this US $4000-odd machine, I ask. “Mostly people over 50,” says Jonathan Weinert, Bosch’s sales manager here. “And for people who live in hilly places and don’t want to get to work covered in sweat.”

Read More

NXP's BlueBox Bids To Be The Brains of Your Car

NXP, the Netherlands-based purveyor of automotive semiconductors, today launched a smart-car computing system called BlueBox that knits together devices that the company already sells.

“We already build processors for radar, vision, and LIDAR, and we see these systems now being combined,” says Bob Conrad, who heads NXP’s automotive microcontroller business.

He spoke from his Austin, Texas, office, where he’d worked for Freescale until last year, when NXP bought it for US $12 billion. That acquisition created what NXP now says is the largest supplier of automotive semiconductors in the world.

Others might regard that claim as a bit of a challenge. Particularly Nvidia. BlueBox can be viewed, in part, as a riposte to the Drive PX package that Nvidia began offering to car makers last year as the basis for the brains of their own systems (notably electric racing cars)

Conrad pooh-poohs Nvidia’s methods: “They have no intermediate processors, and that’s not the way people do it today, not how they bundle options on cars. It’s a Big Bang approach; I’m not saying it will never happen, but it’s not what’s happening now.”

He maintains that NXP’s business has the edge because it doesn’t depend on when or even if self-driving cars become the rule. “About 80 percent of the silicon content is in Level 3,” he says, referring to a standard category of autonomy, in which the car drives itself but the human must always be prepared to grab the wheel (think Google car). “Levels 4 and 5 are only the last 20 percent of silicon content.” 

Level 4 cars let you sit back, tune out, and read a book. Level 5 cars won’t even allow you to intervene: they don’t even have a steering wheel.

Both Nvidia and NXP are putting themselves in the limelight more than usual for companies that supply ingredients instead of a final product. Even their products’ names—Drive PX, BlueBox—slip trippingly off the tongue.

It’s all part of the auto industry’s scramble for the new, data-driven core of the business. Everyone wants to claim dominion over the value-added brains of the car and leave the rapidly commodifying brawn to the OEMs.

It’s all happened before. A memorable example occurred in the PC business, when, back in the early 1990s, Intel broke precedent by advertising as a product what had until then been a mere component. Its “Intel Inside” campaign turned a bland corporate name into a buff, best-selling brand.

Volvo's Self-Driving Program Will Have Redundancy For Everything

Next year Volvo will do something no other company has tried: it will put 100 fully self-driving cars in the hands of customers. The tests, which will begin small and ramp up slowly, are to be held in Gothenberg, Sweden and in London.

It’s a lot harder than it may seem. True, the cars will be in self-driving mode only in the testing area and only when driving conditions permit. But they’ll be production cars, not experimental prototypes, and the drivers will be laymen, not engineers, with full ability to sit back and read a book, not continually putting a hand on the steering wheel to prove they’re capable of intervening at a moment’s notice. There’s no human being to serve as backup.

So far, driver-assistance features have enjoyed the inherent redundancy that comes from having a human being behind the wheel.  “If antilock brakes fail in a car today, the driver steps on the brake,” says Erik Coelingh, a leader of the project. “In a self-driving car, if ABS fails you need a backup ABS.  You need two systems for everything.”

Volvo, a brand with the word “safety” inscribed in it, is insisting on full backup for every element before it allows one of its cars full command. Take the company’s latest active safety feature, an emergency steering system in the 2017 S90 that senses if the car’s about to leave the road and takes control. And, if the motor that’s supposed to twist the wheel should fail, there’s a backup. 

“There are two separate windings—essentially two independent motors in an integrated package,” Coelingh says. “Each winding has its own ECU [electronic control unit], its own power source, its own battery.”

What everybody in the business agrees on is the need for redundant sensing systems. “You have to combine sensors with different physical principles and combine them to compensate for each one’s disadvantages,” he says. “Radar is very much becoming a commodity; camera vision is developing really fast, in part thanks to the development of deep neural networks; and lower-cost LIDAR is coming.”

Volvo’s test cars in Gothenberg and London will each have a small LIDAR unit from an unspecified vendor. The unit will be integrated inconspicuously into the front of the car from where it will scan the road with four beams, unlike the rotating beacons seen on the roofs of Google cars. It will cost far less than the $10,000-plus price of those beacons, also.

High-definition maps of the Gothenberg roads should offer 10-centimeter (4-inch) resolution. Right now the maps are drawn in-house, but when the time comes to expand beyond Gothenberg Volvo will rely on other methods. “In the long run we’ll need crowdsourcing to get up-to-date information on road conditions,” Coelingh says. “We ran a big pilot this winter that used hundreds of cars to measure road friction, sending every reading of slippery conditions to the cloud, so that other cars could share it.”

Such real-world validation is key. “You have to drive quite a lot to make sure that automatic steering doesn’t kick in when you’re not about the leave the road,” he says. “If we get less than one accident in 500,000 kilometers of driving, that’s sufficient.” 

Every step will involve voluminous testing, first under relatively propitious conditions and then progressing through more challenging ones. Volvo hopes the Gothenberg tests will lead to a rollout of a commercial product sometime around 2020, though even that futuristic vehicle won’t work on all roads or under all weather conditions. Not at first, anyway.

And when Volvo does sell a self-driving car, Coelingh says, it will accept full legal responsibility for any accidents that may come. “And that is reasonable,” he says, “because we told the customer he could do something else while the car drives itself.”

How Should a Self-Driving Car Tell You to Take the Wheel?

Someday, the word driver will connote something completely different than it does today. When cars are fully automated and don’t need us for anything more than letting them know where to go, the job of a “driver” will be like that of a patron in a bar selecting a song on a jukebox. But that scenario, where the passenger cabin is a rolling lounge in which all occupants are free to talk, entertain themselves with electronic gadgets, or sleep, is still decades away.

Until then, even the most sophisticated self-driving car will occasionally encounter a circumstance that overwhelms its computerized smarts. At those moments, it will look to a human to take charge. But how will it grab my attention from my phone call or my Netflix movie—or rouse me from sleep—and get me dialed into the potential danger in time to avoid a crash?

Because it’s possible to miss an audible or visual alert if you’re yakking away on a cellphone or engrossed in a car chase scene in a James Bond movie, researchers are looking to add another modality to take-over requests: vibrations in the car seat or seatbelt. A team of researchers at Technische Universität München, in Munich, Germany, and the Delft University of Technology, in the Netherlands, reviewed the spectrum of such alerts, known collectively as vibrotactile displays in the April 2016 edition of the IEEE Transactions on Intelligent Transportation Systems.

Read More

GM and Lyft Will Test Driverless Taxi Service

General Motors and Lyft will test a self-driving taxi service in an undisclosed city within a year, according to a report in the Wall Street Journal yesterday.

It’s not clear what steps the mystery city will take to align its rules of the road with the robotaxi service. Customers who get cold feet at the sight of an empty space where the driver normally sits will be able to opt out of the robotaxi service and get a human-driven Lyft car instead.

Read More

Meet NextEV, the Biggest EV Startup You’ve Never Heard Of

“Nobody has ever done a global startup before, have they?” muses Martin Leach, co-president of NextEV, an EV startup that intends to launch an electric supercar later this year and a mainstream vehicle in 2017. Founded in late 2014, NextEV has its design headquarters in Munich, a software hub in San Jose, and most of its engineers in Shanghai.

“This time last year we were less than 10 people,” said Leach in an exclusive interview with IEEE Spectrum. “Now we’re over 1000, and by the end of this year, we’ll be over 2500. Probably by the end of 2017, we’ll be over 5000.”

NextEV has been raising money at an equally furious pace. Started with money from three Chinese internet entrepreneurs, NextEV was reported last summer to have raised an additional $500 million from the likes of Tencent, Sequoia Capital, and Joy Capital. Now, the company is promising to fund a 10 billion yuan (around US $1.5 billion) research and development effort into electric propulsion and autonomous driving, surpassing a similar $1 billion AI institute from Toyota.

“Our clock speeds are probably five times as fast as a normal OEM [original equipment manufacturer],” says Leach, who was previously CEO of Maserati and president of Ford Europe. The new company has already made a splash in the world of electric vehicles, providing technical expertise that helped propel the NextEV TCR team to victory in the first ever FIA Formula E championship last year. But the main focus of NextEV is an all-electric, road-legal supercar that should be unveiled in just a few months’ time. Contributing to NextEV’s speed in ramping up production is its plan to work to key partners. Says Leach:

Doing everything on our own would be foolish and would take too long. For instance, we don’t have any particular ambitions to get into the business of making actual battery cells. But the way that you put the cells together, in terms of modules and battery packs, is a core competence that we’ll do ourselves.

Around half of the company’s headcount are engineers, working on its power management, motor, drivetrain and digital systems. NextEV just announced that it was a building a 3-billion-yuan ($0.5 billion) factory in Nanjing to produce more than 250,000 motors and electronic modules annually, starting later this year.

NextEV also recently signed a manufacturing agreement with JAC Motors, a state-owned Chinese carmaker. “Our initial focus is on the Chinese market,” says Leach. “China is already the largest car market in the world, but it will be by far the largest zero-emission-at-point-of-use market.”

Part of that is due to the Chinese government’s aggressive carrot-and-stick policies aimed at improving the country’s choked (and lung choking) roads. These include matching subsidies and benefits like access to HOV lanes for electric vehicles, and harsh commuting and license restrictions for internal combustion vehicles.

This is leading to a coming flood of electric vehicles that NextEV’s founder, William Li, admits could disappoint consumers. “The winter of 2017 may be extremely cold for EV industry because of many bad EV products,” he said in a speech in January. “Government subsidies can help to kick off the market quickly, but they can’t make users love EV.”

Leach is adamant that drivers will love NextEV’s supercar—although he admits that few will actually buy it. “The supercar is obviously going to be a very expensive vehicle,” he says. “It’s not done for sales reasons, it’s done for other reasons. The supercar is very much showing our tech prowess, to demonstrate exactly what we can do, technologically and design-wise.”

The plan is for the supercar to provide a brand awareness that, along with NextEV’s Formula E work, might trickle down to the more affordable vehicle in its sights. Bloomberg reported that NextEV’s mass-market car could offer similar performance to a Tesla Model S at half the price. “We’re not trying to be a niche company,” says Leach. “Our products for the larger consumer market will be mainstream products with a premium feel, and we’ll have something towards the end of 2017.”

So far, NextEV has been successful in keeping a very low profile for its passenger vehicles. But one electric vehicle industry insider, who preferred to remain anonymous, says that the company still has a lot of work to do: “NextEV’s charging strategy is not fully baked, which doesn’t surprise me because this next generation of Tesla competitors are over-indexing on designing and producing the car, and under-estimating charging.”

Other experts agree with that assessment. Chelsea Sexton is an EV enthusiast who has advised many car makers. “I looked at hundreds of resumes on LinkedIn and tried to find reasons to find NextEV credible,” she says. “I couldn’t find any. They’re largely an unknown. I have a fair amount of skepticism in general about these start-ups, given that we’ve seen one in the last 20 years of EVs that’s done anything at all.”

Leach remains confident that NextEV can provide an affordable alternative to the Teslas of today and tomorrow. “I don’t worry about the other EV companies,” he says. “The global market for vehicles is over 90 million vehicles a year. Even the combined power of the [existing EV makers] and the new entrants is still only going to be a small fraction of that for a very long time.”

This article was corrected on 3 May to accurately reflect the opinions of a source.


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