|
Getting your Trinity Audio player ready...
|
Autonomous vehicles may be better drivers than their human counterparts, but they seem to get in car crashes more frequently, according to a new study out of the University of Michigan. The study arrived at that conclusion after comparing the reported crash statistics between self-driving cars to national rates.
The study focused largely on Google and the company’s recent intensified testing of autonomous-vehicles in Mountain View, which has tallied more than 1.2 million miles on the road. Researchers also included data from other companies testing self-driving cars, including Volkswagen and Delphi Automotive, but those firms had only minimal testing in comparison.
As of September, Google’s fleet of self-driving cars was involved in 11 crashes, all of which were determined to be the fault of other drivers behind the wheel of conventional vehicles. In a flat comparison of crashes per mile driven, that accident rate for self-driving cars would be more than four times the national rate for accidents, according to the report.
But those numbers needed to be heavily adjusted for a useful comparison, explained Brandon Schoettle, a researcher with the University of Michigan’s Transportation Research Institute. Human drivers often fail to report crashes, whereas any collision involving a self-driving car has to be reported under California law, he said. After adjusting for this and other factors, his new numbers showed self-driving cars had about twice the crash rate.
But that’s not to say autonomous vehicles are more dangerous. Schoettle emphasized that autonomous vehicles have only been involved in car-on-car crashes, and those collisions have always been caused by other drivers. Self-driving cars are rear-ended more often than regular cars, but they are much less likely to be hit head-on.
“The big picture for self-driving cars is even though you might be involved in more crashes, it’s still a safer crash to be in,” Schoettle said. “It’s a mixed message. (Self-driving cars) have higher crash rates, but certainly not because they’re doing anything wrong.”
Asked about the study, Google spokesman Johnny Luu said that the company’s self-driving cars have not once been at fault in an accident. There were no crashes involving the company’s autonomous cars in October, he noted.
“We publish the details of all crashes we’ve been involved in on our website each month, and there’s a clear theme of human error and inattention,” he said via email.
Exactly why human drivers are more prone to crash into self-driving cars is something of a mystery, Schoettle said. Taking a guess, he said perhaps human drivers weren’t used to the driving patterns of self-driving algorithms, which can sometimes result in the cars braking more frequently, and longer, than a human might consider reasonable.
Schoettle emphasized that his analysis had a wide margin of error and it was based off a limited data set. Due to those limits, he cautioned against drawing any hard-line conclusions about self-driving technology. This kind of analysis would become more useful as self-driving cars accrue more miles on the road.
“We’ll continue this line of thinking,” he said. “You can have a vehicle with a higher crash rate that is still overall safer. The two aren’t mutually exclusive.”
-
A winner of the Breakthrough Prize, Berkeley Associate Professor in Mathematics Ian Agol poses on the red carpet at Moffett Federal Airfield on Nov. 8, 2015. Photo by Michelle Le -
Breakthrough Junior Challenge winner Ryan Chester talks to reporters on the red carpet at Moffett Federal Airfield on Nov. 8, 2015. Photo by Michelle Le
“Exactly why human drivers are more prone to crash into self-driving cars is something of a mystery, Schoettle said. Taking a guess, he said perhaps human drivers weren’t used to the driving patterns of self-driving algorithms, which can sometimes result in the cars braking more frequently, and longer, than a human might consider reasonable.”
Yes, from a close call I had with one of these cars, I can say this is likely the reason. These cars drive differently than human drivers. The human drivers on the road are used to predicting the movements of other human drivers. We need to now get used to the movements of machine drivers. In some cases the machine drivers interpret danger where a human driver can tell there is no cause for concern. I have seen this with these cars several times and now I’m getting accustomed to driving around them. I think it just takes practice for people. It’s not necessarily about good or bad driving, it’s just different and it takes some time for people to adjust to the different movements.
I have noticed myself that when I am driving near one of the Google cars I am very aware of it and I get the feeling that other cars are a little more wary too.
That being said, I was riding as a passenger with a local and we stopped behind a Google car. I made some comment and the driver was not aware that it was driving itself which was quite interesting since she drives the area a lot. If other drivers are not aware for some reason that this is a self driving car do they behave differently than if they are aware? I had not noticed this myself.
On another occasion I was stopped at traffic lights in front of a Google car and the car beside me contained several people with arms out of windows taking pictures and videos of said Google car. I wonder if some drivers are looking on the Google cars as some type of tourist attraction that needs to be observed rather than paying attention to driving themselves.
Yes, they have different driving habits…they follow the laws. Those not following the laws crash into them. Sure is a mystery. I hope they solve it.
They take a long time to get going at a four-way stop. I can see someone rear-ending them if they were expecting the car in front to move more quickly.
Yes, if a driver is not paying attention to the car in front of them, ans assumes they will move forward, and then act on that assumption with acceleration while not giving a safe enough distance, they are bound to run into other cars.
Maybe Google can start working on self driving bicycles.
People generally aren’t very cautious drivers. You might be able to get by with below-recommended following distance in normal situations, but any kind of hard stop or longer-than-expected braking and you might be rolling the dice.
I’m hoping that automatic braking (when your car detects there’s a high probability of hitting the car ahead of you) will be a standard feature on all new vehicles — especially for human drivers — over the next 5 years. Because if you don’t have it, your insurance company will be unhappy writing a check if you rear-end the fancy bumper filled with sensors on the self-driving car in front of you.
What I have noticed. Does driving 5 to 10 MPH slower make you a better driver? The other drivers are always passing the Google car because they are the slower vehicle often driving under the speed limit. They also brake too often under some fear someone will turn or a person will walk out in front of them.
Aaah, hahaha. Having to deal with cars programmed to drive within the laws seems to be something difficult for some to adjust to. Proof of just how law-less people act in their cars on a regular basis.
“Quick,someone complain about bikes…we’ve been exposed!!”
Self-driving cars are still highly experimental, so it’s impossible to draw any conclusions about how they will perform 10 years from now when (if) they turn into real products.
This article is the automotive equivalent of: “Cold fusion reactors: cheap fuel but not stable enough.”
The study uses the national rate for accidents, but this is Mountain View, a three by three mile stretch of land with city streets. How can this analysis possibly work out?
Also, where is the link to the original study, Mr. journalist? I know I can just Google for it, but it’s tedious, with the name of the author down in the middle of the article and neither a specific title or where it was published. Please fix, thank you.
I hope these cars can hit some illegal immigrants and send them back to mexico!
SLOW .. 25mph is about the speed of a moped
The prototypes’ speed is capped at a neighborhood-friendly 25mph, and they’ll drive using the same software that our existing Lexus vehicles use—the same fleet that has self-driven over 1 million miles since we started the project.
http://www.dailymail.co.uk/sciencetech/article-3314592/Google-adds-cartoon-painted-doors-self-driving-car-bid-make-scary.html
The sooner we can get the average inattentive MV driver into an autonomous car the better. Then those dullards will quit running into them, pedestrians, cyclists and train crossing barriers thus leaving the roads safer for those of us who know what we’re doing.
Old News – was reported months ago that Google car crash rate was much higher than average.
And while adjusting for reporting bias, it does not sound like the report adjusted for traffic/speed bias. That is, Google cars mostly operate on very slow and safe suburban streets, which means you really should adjust the Google rate upward.
It also fails to factor in the impact Google cars have on other drivers – the Google cars are so slow and tentative that they cause other people to have accidents – not necessarily into Google cars.
Google has made huge tech strides in self driving, but it is still like handwriting recognition in the 1990s – more of a problem than a solution so far.