Google Inc has issued the first of its monthly reports on incidents with its driverless cars after the second accident involving one of its vehicles occurred within a week.
Following initial refusal at a shareholder's meeting last week by the company's co-founder, Sergey Brin, to publish incident data, Google said on Friday it will now do so on a dedicated website.
The first report shows an autonomous vehicle stopping to avoid two cyclists crossing a junction, one of whom was apparently salmoning (riding against the flow of traffic) towards the car. Since May 2010 Google Inc reports 11 collisions with its cars, four of which happened in 2015, and none of which were blamed on the autonomous car.
While Brin wants Google cars, which could be tested on city streets from this summer, to "beat human drivers", believed to be responsible for 90% of crashes, the company has attracted criticism for a lack of transparency by withholding incident data during years of testing the vehicles on the streets of Mountain View, California.
Consumer Watchdog's Privacy Project director, John Simpson, who confronted Brin at the shareholder's meeting last week, told USA Today:
"Google is dribbling out bits information in the hope to silence legitimate calls for full transparency,"
"They are testing on public roads and the public has a right to know exactly what happened when something goes wrong."
Simpson has previously said Google Inc eventually wants to do away with the steering wheel and pedals altogether, so a person cannot take over if the vehicle runs out of control, making it even more important crash data is made public.
There were 11 minor accidents involving Google's autonomous cars in six years and 1.7m miles of the project's testing in Mountain View, 1m of which were during self-drive mode. One of Google's autonomous vehicles was apparently rear-ended at a red traffic light on Thursday while last weekend another unit was hit by a vehicle.
The first report also shows situations of interest, including one at night (pictured above) where the vehicle kept track of two cyclists, one of whom moved into a left turn lane before the cyclist "veered back onto our path to continue straight across the intersection", according to the report.
The report goes on: "At the same time, the cyclist on the right entered the intersection, traveling against the flow of traffic. That cyclist then took a sudden left turn, coming directly at us in our lane.
"Our car was able to predict that cyclist’s path of travel (turquoise line with circles) so we stopped and yielded. This happened at night, when it would have been very difficult for a human driver to see what was unfolding."
Proponents of the designs argue their radars and sensors make the cars able to understand their surroundings better than human drivers, as well as having faster reaction times. The cars can honk horns and flash at distracted drivers, and tighten seatbelts and take evasive action if they sense a likely crash.
As well as data on crashes, examples of how the cars react to common traffic situations will be published. For now Google cars are still accompanied by a human driver.
The company said: "We've made a lot of progress with our self-driving technology over the past six years, and we're still learning,"
Add new comment
25 comments
I'm amazed they managed to hack self-driving car code out of the 1980s Star Wars arcade game. SpaceX publish HD footage of their return-to-base rockets blowing up - video or it didn't happen.
So if you do manage to get yourself run over by a Google car and it wasn't your fault; do you sue Google, the person sitting in the car, the owner of the car, the manufacturer of the car, the manufacturer of the sensor / chip /hardware that went wrong or the computer coder?
I wish we'd scrutinise human drivers as much as we scrutinise Google cars.
[[[[[[[[[ Er, no, the comment beginning "As a cyclist that puts...." comes from "Forcrz", and not from me. Dunno what happened there! P.R.
Last night on a ride I turned right out of a junction, with about 100 yards to go before I had to pass between some traffic islands (https://www.google.co.uk/maps/@53.572028,-2.494255,3a,75y,180h,90t/data=!3m4!1e1!3m2!1srV3kV-opyH2vtFGRzTE7Kw!2e0!6m1!1e1). A car followed me out of the junction and then, just as I was approaching the traffic islands, accelerated past me with a great noise and nipped in between the islands in front of me. Fortunately I've come to expect this sort of idiocy, so had pre-emptively slowed down and was hanging left, but even so. I threw up an arm in exasperation and got the wanker sign and the finger out of the window as he shot off. Would a driverless car have done the same? No, so bring it fucking on.
I was thinking about the very same thing on Sunday when a few impatient idiots wheezed past me just because they think they are about to loose a few seconds. Computers - in theory - would want to drive over me just because they are in a hurry.
bring on johnny cab
We should remember that cycling is one of the few sports/activities increasing in popularity, we are not going away and car drivers have to understand that/get used to it.
As a cyclist that puts 2 to 3k of miles a year it is the tools like these that give cyclists a bad name. Not to mention the idiots that try to justify it. Look at the data, the idiots were in the wrong lane on the turn and in the wrong lane period. I can care less how you try to justify it. They were salmon and ignorant. All cyclists need to start following the law and rules of the road. Stop justifying the action of the idiots. We need to improve our name. And sooner or later they will approve licensing and reg for cyclists because of evidence that shows it is our fault and not the cars.
Was it a slow day over on the daily mail web site so you come here with some rob liddle bile. Motors need to improve there name, there the ones kill and injuring thousand of people year.
OK I'll take the bait. th
How and where has licencing and registration stopped drivers behaving like complete tools? It hasn't drivers still drive like tools.
How would licencing and registration prove someone was at fault?it doesn't
80% of adults ridding a bike have a current driving license. Regular cycling probably makes most of them more alert driversdrivers. It does me. Do you think cycling is so demanding it needs extra tested skills over and above driving? If then you are doing it wrong.
Doubtless you will be in favour of all children receiving cycling training at school and practical cycling to be an essential part of the driving test.
Please go back to the daily mail
Problem is that cyclists are an out group, and I believe there are enough moaners who would put the effort into campaigning, that driving licences don't stop stupid behaviour isn't actually relevant to the argument. You are talking about people who simply do not want cyclists on the road, who do not want anything to get in their way. These are the "perfect" drivers, the ones who never speed unless it is safe, never ignore traffic lights, unless it is safe etc etc.
What evidence would that be? Your confirmation bias filtered anecdotes do not count as evidence.
This does, and shows that three out of four bike/car collisions are down to the driver. It is from the Transport Research Laboratory, and was reported on this site.
"Bike riders tearing through red lights, wearing dark clothing or riding at night without lights are to blame for less than 7% of accidents that result in a cyclist being seriously injured, according to research commissioned by the Department for Transport.
The study, carried out by the Transport Research Laboratory – which has also published a report on helmet-wearing that we have covered separately today – found that one in four accidents resulting in death or serious injury to a cyclist was due to the bicycle being struck by a vehicle from behind."
Meanwhile, according to police reports studied as part of the research, wearing dark clothing at night was thought to be a possible cause of just 2.5% of accidents resulting in serious injury to the cyclist, with not using lights or jumping red lights each blamed in 2% of cases. Those percentages rose slightly in instances when the cyclist was killed, although in those circumstances police could only rely on evidence from the driver and other witnesses.
"The report’s findings show clearly that far from being the danger to other road users that certain elements of the media have portrayed them as in recent months, cyclists are themselves put at risk by the actions of motorists, with the police attributing blame to the driver in up to three quarters of collisions between a bicycle and other vehicle in accidents involving adult bike riders."
http://road.cc/content/news/12065-report-dft-casualty-stats-says-cyclist...
An old saying:
"To err is human, but to really fuck up one needs a computer"
Automated cars are the slippery slope - Automated bikes anyone ? Way too dangerous and controlled by humans, who are fallible.
I think Wallace and Gromit were on the right track. Automated techno trousers for everyone, it's the only way to be safe.
Oi! My 4x4.
Sweeping generalisations are only bad when they are used against cyclist.
I'm hung like a horse too.
Yeah, bring it on, I'll take my chances with a google car, rather than your average white van man. ...... Or small membered 4x4 driver
This can’t come soon enough. They’re going to better drivers than most humans.
Once these vehicles become widely available courts will no longer be able to use the loss-of-livelihood excuse for not banning dangerous drivers because there will be an alternative.
Bring it on.
Can we start replacing minicabs with Google cars now? Even given the concerns surrounding automated cars, I think I would feel more comfort on a bike around them then the average London minicab in commuter traffic...
Not entirely sure I buy Googles reading of the incident given.
Cyclist 1 suddenly jinks right to continue through the junction?
Cyclist 2 attempts to commit suicide?
Are we sure that the Google Car was not just had a glitch and that
Cyclist 1 was always in the lane to go straight-on, and
Cyclist 2 is on the footpath.
Without more information which scenario is more believable?
I think id trust 10,000 times a second laser scanning over erratic city cyclists.
Cyclist 1, took the wrong lane, swerved back into original lane, tell me youve not seen that a million times from road users, be it cars, motorbikes or cyclists.
Look where cyclist 2 came from, the left hand lane of the road, riding against the flow (probably in the margin/gutter), the google car applied a possible path of travel for the cyclist which might intersect and therefore stopped.
Option one.
Now that Google are taking full control of and responsibility for the driving, I'm spending next year's insurance money on a new bike.