After one of Google’s driverless cars recently ran into a bus, the director of the project has highlighted a separate incident involving a cyclist riding on the wrong side of the road as an example of how its technology has in other cases been superior to a human driver.
Inverse reports that a self-driving Google Lexus collided with a bus on February 14. The Google car was stopped in the right lane where it was being held up by sandbags surrounding a storm drain. The bus was in the middle lane when the Google car pulled out.
It seems the bus driver thought the Lexus would stay put, while Google’s driverless car thought the bus would slow down and let it out. The member of staff behind the wheel in the autonomous car also felt the bus driver would give way and didn’t intervene.
Reviewing the incident, Google said: “From now on, our cars will more deeply understand that buses (and other large vehicles) are less likely to yield to us than other types of vehicles.”
Last June, Google reported that there had been 11 minor incidents involving its vehicles since May 2010, none of which were blamed on the autonomous car. Typically, they were hit from behind in slow speed crashes by human drivers unaccustomed to other road users moving with such caution. To address this, the firm has since said that it is working on making its cars more assertive.
Chris Urmson, director of the project, described the recent bus collision as ‘a tough day’ but said there were other incidents to weigh against it.
One he cited involved a cyclist on the wrong side of the road almost running into a vehicle. There had been another cyclist near the car, as well as other drivers and a pedestrian. The car was able to track each of these meaning it was able to stop in time. “I am convinced I would have hit him,” said Urmson. “I can’t shard my intelligence.”
25 thoughts on “Google defends driverless car project after collision”
Having seen the incident. The
Having seen the incident. The Google car was indicating to change lanes in front of the bus. The bus driver continued without slowing, they collided at slow speed. There was a video camera in the bus cab, guess what the bus driver was holding and eating a sandwich at the time. Hardly paying full attention to the slowly unfolding situation!
The bus driver, if he had been a little more patient could of let the Google car move lanes. Not so easy to apportion blame solely on the Google car.
CXR94Di2 wrote:
“Google’s driverless cars recently ran into a bus” “Boom headline”
Sadly the article still start with this controversial sentence. Why do you do that Alex? Personally I think that’s a shitty practice and makes me less likely to read road.cc.
As I’m also in the bay area, I bike near various autonomous vehicules every now and then albeit I never had anything special happening, until – coincidentally – 3 weeks ago in SF.
I saw a new one called something like “Car Kit” which I’ve never seen before. Zillions of sensors you apparently bolt onto the car. As a software engineer, wasn’t too sure the algorithms of this would be nearly as good as Google’s who’s been working on it for so many years, even thus as I bike in the city every single day for years I also figure any program is going to be more likely to be attentive and have basics dialed in than a human on his phone (90% of drivers in SF, seriously – not that the bike riders are any better).
I look through the window and the human driver is not looking at me (hes actively driving it though at this point) – but the car is slowing down and I suspect its going to turn in front of me so i’m cautious.
5m away from me the car suddenly stops a good distance before the crossing (further away than any human ever does). I was looking at the driver the whole time and only when the car stops did he look in front of him and at me – a little concerned.
Thank you, car software, I think, as I cross safely on my bike.
..the fact that it actually
..the fact that it actually noticed the bus and indicated before it pulled out is an advance in my book…
Shard is now a verb?
Shard is now a verb?
#bloodykidstoday
Dr_Lex wrote:
In the quote in the article, can someone please translate what he said, cos I don’t understand what he means… “I can’t shard my intelligence ” ? – wtf?
brooksby wrote:
Sharding is a computer process involving the partition of a hard drive so that data can be retrieved more quickly. (I looked it up). If the driver works for Google and if he is an IT guy this may be normal language for him.
WiznaeMe wrote:
OK, great- thank you 🙂
I feel a bit uneasy about the
I feel a bit uneasy about the whole concept of driverless cars but I reckon it will basically be OK because the people most likely to buy them will be the sort of people who have no interest whatsoever in cars or driving. A computer is therefore more likey to be a better driver than that disinterested person!
Just my 2p worth.
rollorawles wrote:
It’s not the uninterested driver who worries me, it’s the bloke who is passionate about cars. Anyone who spends more than ten percent of his disposable income on his motor and whose back wheels have been modified away from the vertical is, by definition, a menace. He ( it’s always a he ) invariably lacks respect for other road users and exhibits a breathtaking ignorance of the laws of physics.
Only yesterday I passed a Subaru that had skidded on well forecasted ice, punched a ten foot hole in a farmer’s hedge and ricocheted fifty yards into the field on its roof. The most prominent component of the upended vehicle was the shiny new after-market exhaust, with it’s five-inch diameter tail pipe. God protect us from the heart-stopping incompetence of the enthusiast.
mike the bike wrote:
This.
Those that see cars as transport are far more likely to be safe drivers than those that drive for fun.
rollorawles wrote:
Don’t forget that some people aren’t allowed to drive. I had a TIA (mini stroke) 6 years ago, leaving me with a blind spot which means I can’t drive. I suspect that when the same thing happens to somebody in 10 or 20 years time they’ll be able to carry on using their motorised transport instead of having to rely on unreliable public transport and Shanks’s pony.
Yes, I do cycle, but sometimes you need to arrive dressed smartly and not sweating like a pig. (E.g. As Stanley Holloway almost said, you can’t go to a funeral in cycling shoes. http://www.goodfuneralguide.co.uk/2012/07/brahn-boots-stanley-holloway/ 🙂 )
rollorawles wrote:
I’m a disinterested driver according to your definition; I don’t do it very often and I honestly can’t see why anyone would choose to do it, especially in town.
On the other hand, I’m not totally desensitised to what I’m doing (steering a big lump of metal through the streets) and I don’t think I’m God’s gift to driving. I think these are characteristics of ‘interested’ drivers and I’ve no doubt a computer is likely to be a better driver than most of them too.
Thanks – in the context, I
Thanks – in the context, I took it as nerdspeak for “I couldn’t evaluate all the data and make a decision as fast as the AI can”, but was stuck with the traditional meaning of shard as a fragment rather than a possible acronym of ‘System for Highly Available Replicated Data’ (having looked it up).
Every day’s a school day…
Shard in this case is
Shard in this case is probably more likely to be a reference to the practice of breaking up game servers for thousands of players into multiple parallel game spaces which look identical but contain fewer players each. In which case it’s a bit of a nerdy way to say “I can’t multitask” which used to be a nerdy way of saying “I can only concentrate on one thing at once” but has been adopted into the mainstream by the Muggles 😉
Perhaps.
Morat wrote:
Almost, it’s not “I can’t multitask”, it’s “I can’t separate the tasks”. Sharding is a term now used all over the place in computer science and IT, but its original use is in database design as applied to building large scalable systems. A design that is difficult to shard, is one with many interdependencies between components.
It might also be a reference to machine learning, and dividing (sharding) the problem to the level that current neural net algorithms can be trained to resolve.
Sounds as if to a large
Sounds as if to a large degree it wasn’t really the Google-car’s fault, if the bus-driver wasn’t properly paying attention. But there are lots of things about this that make me wonder.
What it seems to emphasise is that a self-driving car in the presence of human-driven vehicles has to make judgements about what the human is likely to do. Hence ” our cars will more deeply understand that buses (and other large vehicles) are less likely to yield”.
But isn’t there going to be a feedback loop when driverless cars are common? If drivers of larger vehicles learn that self-driving-cars will behave submissively to them, they’ll probably become still less likely to yield to them, really taking liberties, in fact. Hence requiring the cars’ reactions to be retuned to be more passive still.
And the purchasers of self-driving cars might in turn get frustrated at this and start demanding more aggressive software (or installing hacks?).
Just seems to me that once on the road in significant numbers its going to be a whole different game (and perhaps game theory is the relevant discipline?), between human drivers and presumably multiple different providors of auto-car software (including hacked or even Open Source?)
The video of the accident
The video of the accident doesn’t seem to be available, and many of the descriptions of it lack detail.
But apparently(according to Google’s report) it happened in a very wide lane where customarily cars turning right hug the curb to allow others to pass. But the far right of the lane was blocked by some sandbags, so the Google car had to pull out to avoid them, into the path of the bus. No mention if the Google car indicated, but pretty obviously it was stopped and then moved into the clear travel area and into the side of the bus. 99% the fault of the Google car.
Meanwhile today about 2-3,000
Meanwhile today about 2-3,000 people have died in car crashes caused by human failure.
Yes, that’s right 2-3,000 people dead (1,000,000 / year world wide roughly) because of human failure while newspapers & readers concentrate on a minor bingle by an autonomous car.
I love driving myself but I look forward to autonomous cars so that all the dumb drivers who fail to concentrate at the wheel can hand over risking my life to a computer. (As long as it’s not controlled by something like Microsoft Windows 95 etc. with their “bue screen of death” for those of you old enough to remember that!) Autonomous cars may not be “perfect” but they are already a darn sight better than most drivers on the road. I expect & accept that there may be the occassional stuff up by computers but safety for everyone on the roads will be still be better than for human drivers.
Hackers or political control freaks could be a problem though (ala Atmos in Doctor Who.)
Quote:
Wait a minute, are you saying that Google is in a conspiracy with the Sontarans? I always felt that they’d work better with the Great Intelligence…
You can be certain that
You can be certain that driverless cars WILL be safer than human driven vehicles and by a very wide margin because they would never be approved if they were only marginally safer.
If they are aiming for 5% of current level of accidents they will achieve that. They will not aim for 75% of the current level.
Once they are in use pressure on owners to switch to them will soon grow. The first question a driver in a collision will be asked will be ‘why did you choose to drive when figures show it is 20 times more dangerous?’. Fewer drivers means much higher insurance premiums. Courts will be much more likely to hand out the bans they should be giving now when no-one needs to drive.
ooldbaker wrote:
Will motorists receive a financial bung to trade their car in for a new self-driving electric model? Most motorists hang onto their car until it can’t drive, few just trade them in on a whim.
brooksby wrote:
I imagine that the insurance premiums will rapidly reduce once the safety/efficacy of driverless cars has been established. So, yes, drivers will have a financial incentive to get a driverless car.
hawkinspeter wrote:
And insurance has the greatest impact on heavy mileage users. Cars have a lifetime of about 15 years, but heavy users upgrade sooners and vehicles get passed to mostly low mileage users. With the insurance incentive, the shift in terms of their share of total journey miles (modal share) may be quite fast.
Another point I do like to raise regarding self driving cars, is that the same technology does open new ways to deal with dangerous drivers. Barely half of drivers receiving 12 points lose their licence, pleading “excessive hardship”. That excuse is about to become (even more) ridiculous.
bikebot wrote:
Yep, I’m looking forward to when most motorised vehicles are autonomous. However, we may see even more congestion on the roads unless there’s a big change in shared vehicle usage (e.g. buses).
In my view of the future, the space given over to car parking will be largely reclaimed. When autonomous cars become useful and cost effective enough, people won’t need to own a vehicle and have that vehicle spend most of its time parked up not being useful. Instead, a smaller number of shared vehicles will be able to service the same number of journeys and out-of-town parking can become feasible for non-peak times.
“From now on, our cars will
“From now on, our cars will more deeply understand that buses (and other large vehicles) are less likely to yield to us than other types of vehicles.” –“From now on, our cars will more deeply understand that buses and other large vehicles are also driven by inattentive arseholes.”