Germany, US lead research into robotic cars

As the American state of Nevada legalizes autonomous vehicles, researchers indicate that German and American corporate money is behind most research into driverless cars.

Just last week, after some lobbying from Google, the American state of Nevada became the first region of the US to authorize autonomous cars on public roads. The Nevada Department of Transportation is now tasked with adopting "regulations authorizing the operation of autonomous vehicles on highways."

Google revealed last fall that it had been driving its driverless car on Californian roads over 225,000 kilometers (140,000 miles) for many months in 2010. That new law comes just after the world's top researchers into these robotic cars gathered at a recent conference in Baden-Baden, in southern Germany, to share their latest results.

The research into autonomous driving was kickstarted seven years ago, when the US Department of Defense research agency, better known as DARPA, created a desert road race for self-driving vehicles.

The race was tough - not one of the 15 competitors made it to the finish line. Things have vastly improved since then though. Today, kitted out with video cameras, radar sensors or laser range finders, test cars are able to drive themselves through real city traffic.

"It is a very surreal experience at first because you are sitting in the driver's seat and so you are so used to having to control everything," said Jesse Levinson, a researcher from Stanford University, who attended the Baden-Baden conference and presented a paper.

"And then all of a sudden, the steering wheel is turning by itself and the gas is being pressed by itself. And all of a sudden, you start getting comfortable and you say, 'Hey, wait a second, my car is driving and it is actually doing a really good job.'"

Alberto Broggi (left) led a project to drive a car from Italy to China in 2010Levinson added that autonomous cars with video cameras, computers and laser sensors, are now better than humans at doing certain things on the road.

"It's even more accurate in staying in the middle of the lane, and it is able to detect when a car in front of you brakes, even faster than a human driver," he told Deutsche Welle. "It also has a better sense of what's going on behind itself than a person."

Autonomous cars not perfect yet

Despite their high-tech equipment though, self-driving vehicles still fail dismally in certain situations. In extremely heavy rain or when there is a lot of dust for example, the cameras just can't see the road. It is possible to circumvent this with special sensors, called laser radar sensors, or ladars, in short.

Of course, ladar doesn't actually "see" anything, rather it scans and processes the echoes bounced off targets. But ladar brings difficulties of its own, explained Alberto Broggi, a computer scientist from the University of Parma.

"It is getting pretty common to have a one of these quite expensive sensors on top of your car that is able see everything around you; the front, back, the sides, very far away," he said. "The problem is to process this data. Because it is a huge amount of data. And you have to understand every single echo from this system and classify that."

Last year, Broggi's research team tested an autonomous vehicle over a staggering 30,000 kilometers (18,600 miles) , taking it from Italy to China. Broggi says that the logical computers driving the car were completely overwhelmed when humans drove chaotically – a common scenario in huge and sometimes unruly cities like Moscow or Shanghai.

Broggi said that Moscow traffic frustrated the autonomous driving softwareHe noted that's why artificial intelligence researchers like Broggi are now focusing on improving the way the data from the sensors is processed. But, the data itself can be overwhelming.

"We secured some 20,000 hours on the largest supercomputer in Italy, so now we have a huge amount of data because we collected every single bit that came out of the sensors and from GPS and from radios," he said.

Broggi expects to spend the next three years at his desk crunching numbers.

"The idea is to process the data, find out what didn't work, then go back to the laboratory and fix the issues, fix the software, make changes and then process it again, and then you hopefully get a smaller amout of data that doesn't work, then we will start working on that again," he said. "And then we will process and process again the same data until we get some software that can process everything 100 percent reliable."

Elements of driverless robotics are already available

While completely autonomous cars are still in the testing phase, some spinoffs can already be found in commercially-available vehicles.

Autonomous cars are designed to stop for pedestriansMany German carmakers such as Daimler, BMW, Volkswagen and Audi as well as companies associated with the automotive industry, such as Bosch, have been quietly pushing this type of research in Europe. With some serious money being thrown at the technology, as well as backing from Google which is testing its own fleet of autonomous cars in the United States, many experts predict it won't be too long before self-driving vehicles hit the streets.

At the intelligent vehicle conference, Joachim Happ, a Mercedes engineer, who took people on a drive in a simulator showing them so-called driver-assist technology that is already available in top-end Mercedes cars.

"My desire is to drive fast, for example, at 150 kilometers an hour, but there is a car ahead of me only driving 120, then it autonomously reduces the speed, by throttling the engine, or if necessary even braking by itself," he explained. "If the driver unattentively leaves the lane, we already have an algorithim pushing him or her back into the desired lane."

Author: Kate Hairsine, Baden-Baden / cjf
Editor: Nicole Goebel

http://www.dw-world.de/dw/article/0,,15194126,00.html