Automated Vehicle Conference Speakers Foreshadow Impending Driverless Future

Andrew Chatham, Google. AUVSI photo.

By Danielle Lucey



The leading self-driving vehicle experts speaking at this week’s Automated Vehicle Symposium 2014, held by AUVSI and the Transportation Research Board in Silicon Valley, California, discussed a not-too-distant future where vehicles could pilot themselves, and the timeline for integration closer to now than ever before. 



Speakers from Daimler, BMW and Nissan all discussed their companies’ future plans in the automated driving arena, with each putting the technology on the roads within the next five to 10 years. 



Dr. Ralf Herrtwich, the director of Daimler’s driver assistance and chassis systems, discussed the company’s two-prong interest in automated driving, both consumer vehicles owned by individuals, which can be implemented with lower levels of autonomy, and car sharing endeavors like Car2Go, which would require a higher level of autonomy to make worthwhile. 



“The thing is you have to understand the economics of this car sharing business. The costs that you incur running such a business are mainly driven by the number of vehicles you want to employ in the field,” he said, noting it’d be ideal to have so much autonomy that the cars would pilot themselves to the customers that desire to drive them.  

 

Herrtwich also addressed the safety factors of increasing autonomy in automated systems. 



“Automation and safety, that’s a tricky relationship,” he said. “It’s a good success story so far, because we’ve been seeing in all markets … the numbers for fatalities and injuries coming down in the last years.”



He said automakers have to deal with the reality, however, that humans do many things correctly when driving, and the number of errors they make are small by comparison, although those scenarios might be the cause of many accidents today. 



He also warned that automakers, like Daimler, which makes Mercedes-Benz, must persist with automated driving to survive. Herrtwich showed a slide displaying the names of about 10 horse-drawn coach makers that no one in the audience was able to correctly identify, and that could be the future of an automaker that doesn’t evolve, he said. 



“If there is something more than this, we at Mercedes want to be part of it, you can be sure of that.”



Ogi Redzic from HERE Nokia, the company’s mapping software, and Andrew Chatham from Google’s self-driving car division, discussed the digital infrastructure the two companies use to let vehicles map out their path. 



“We know that in order for us to be able to provide maps that will allow automated vehicles to drive by themselves, we’re going to have to start from a very different starting point,” said Redzic, explaining that typical mapping is not going to be enough. The company uses lidar technology to create maps made of millions of data points, which he says would be key to creating top-notch mapping for self-driving. 



Chatham concurred, saying many people incorrectly assume that the map the Google self-driving cars use is the same as its Google Maps app. 



“A map to us is any sort of geographic information we can tell the car about in advance,” he explained. 



The company images roads, collecting data and continuously adding information for the cars to interpret. 



“If I had shown you the same intersection two or three years ago on the project, it would have been much simpler,” he said, and even with those rigorous changes, things like road construction can alter the terrain all the time. 



Some speakers did caution of the difficulties of the task at hand, like MIT Prof. John Leonard, who participated in the DARPA Urban Challenge. His team’s car overcorrected on a passing maneuver in the competition and had a collision with Cornell’s SUV. 



“We had about five bugs, and they had about five bugs, and they interacted to form this incident,” he said. 



Situations like this have made him preach safety above all else to his students since then, he says. He also cautioned that the technology challenges that lie ahead are large. 



“I do believe in the vision, … but if we really need to interpret human gestures and social cues, those are some of the hardest artificial intelligence issues I can think of,” he said. 



“I think it’s hard to convey to the public how hard this is, and it’s really hard.”



Presidential Announcement



The first day of the Automated Vehicle Symposium, which wraps up this Thursday, coincided with an announcement from President Barack Obama on the future of the United States’ driving infrastructure. 



Today he urged Congress today to address vehicle safety in the country through his newly proposed $302 billion, four-year transportation plan, which addresses adding vehicle-to-vehicle and vehicle-to-infrastructure technology to America’s roadways. 



Obama made these remarks at the McLean, Virginia, Turner-Fairbank Highway Research Center, where he had the opportunity to test out a simulator equipped with the technology. 



“They can tell you if an oncoming vehicle is about to run a red light or if a car is coming around a blind corner or if a detour would help you save time and gas,” he said. “And I got to test all this in a simulator. It was sort of like ‘Knight Rider.’”



Obama cited a study that said Americans spend 5.5 billion hours each year in traffic, which costs $120 billion in wasted time and gasoline. 



The president expressed frustration in the partisanship that has been blocking infrastructure updates from occurring in the country. 



“This is not an abstract issue,” he said. “And it shouldn’t be even a partisan issue.  Republicans, Democrats, independents — everybody uses our roads.”



The current two-year highway bill is set to expire in November.