Even the most experienced drivers would be proud of this statistic: 700,000 miles without an accident. Things happen. Other drivers are overly aggressive. Weather conditions make driving unsafe. People drive when they are really too tired or too emotionally distraught to do so. There are impaired drivers making things even more challenging. Even with all of these variables, the new Google self-driving car software has logged over 700,000 miles of accident-free driving while commuting in both city and highway conditions without the assistance of human drivers. That is a perfect track record over a distance that would more than cover 117 round-trip commutes from Los Angeles to Boston.
Back in 2009, when engineers first publicized plans to test the Google self-driving car software, they focused primarily on highway driving. Over the past year, however, this has changed. Testing engineers believe that the Google self-driving car is close to achieving complete mastery on the highways. The last year has seen testing on the automated driving technology on the streets of Mountain View, CA, where the info-tech giant is headquartered. It made sense for Google to start testing on the freeways, where pedestrians, stoplights, cyclists and confusing intersections do not exist. Now the company believes that its technology is ready to navigate through all of these precarious situations and more.
“[Google has] improved [the] software so it can detect hundreds of distinct objects simultaneously…” Google advertised in a blog post. “A self-driving vehicle can pay attention to [numerous] things in a way that a human physically can’t—and it never gets tired or distracted.”
Most people would feel a little leery trusting their safety to a fully automated control system, considering the countless variables that come with city driving. So far, however, Google’s self-driving car has not hit any roadblocks with its city testing and, in fact, still has a perfect track record in terms of the number of accidents that its vehicles have been involved in. Just like the highway tested vehicles, so far the city tested Google self-driving cars have not recorded one accident. Google is confident that the automated self-driving software can perform even better than humans in both highway and city driving exercises. The company says that the software can fully detect pedestrians, cyclists, construction zones and even hand signals and respond accordingly.
“…What looks chaotic and random on a city street to the human eye is actually fairly predictable to a computer,” said project director Chris Urmson.
Of course Mountain View, CA with its reasonable suburban traffic and its wide streets does not pose the hazards that Chicago, Oakland or New York City would, and engineers readily admit that the Google self-driving car needs more testing and some tweaking before it would be ready to navigate these heavily congested areas. Even though the automated robot is now programmed with “vision” to “read” road signs rather than relying on a pre-programmed map to tell where and what they are, as well as the ability to tell the difference between hundreds of different objects and people in real-time, there are still some issues. The Google self-driving car has had problems with changing lanes, merging and turning right on a red light. Also, bad weather has been a challenge for the software.
“We still have lots of problems to solve,” Urmson said. “…But thousands of situations on city streets that would have stumped us two years ago can now be navigated autonomously [by the Google self-driving car].”
Google is holding fast to a timeline that co-founder Sergey Brin predicted back in 2012. The company believes that its self-driving software will be available for public use by 2017. They say that at that time the initial software will probably need drivers to be ready to take the wheel in the case of a glitch or computer failure. The long-term vision of the product, however, is to eliminate the need for anyone in the driver’s seat at all.
By Jeremy Mika