Slashdot Mirror
Google's Autonomous Car Passes 2 Million Miles
Google said today it recently reached a huge milestone with its autonomous cars. Its cars have logged its two-millionth mile. To put into perspective, Google's self-driving cars have travelled roughly 300 years of human driving. The first million miles took Google six years, the second million came in at 16 months. Recode evaluates how far Google's self-driving cars have come. It notes that Google has been involved in 14 of such incidents, 13 of which were caused by other drivers.
As I understand it, the test cars don't go faster than 25mph. After so many miles of testing, will the car finally be ready to drive a bit faster?
"Google's self-driving cars have travelled roughly 300 years of human driving."
Quantity != Quality. 300 years of human driving involves a lot more than driving 25 mph on suburban roads in ideal weather conditions.
They won't get much for that car when they sell it, it's hard to sell a car with more that 100,000 miles on it.
I was just in Mountain View, and saw a Google car stopped at a traffic light. The light turned green, but a pedestrian jumped out in front of the car anyway. The Google car just sat there until the pedestrian left the crosswalk.
"It notes that Google has been involved in 14 of such incidents"
14 of _which_ such incidents? I mean, i can make a pretty good guess, but if i were to read the blurb without any context i might think this is the 14th time Google has passed a million miles or something.
This Space Intentionally Left Blank
Tesla has collected over a billion miles of driving data (about 180 million on autopilot) and adds a million miles every 10 hours.
I don't read your sig. Why are you reading mine?
So, if I get this right, those Google cars cause about 0.5 accidents per 1M miles? If so, that equates to about 1.5M traffic accidents per year in the US if you replaced every car with a driverless model (assuming all rates are constant, of course). If that seems like a big number, Americans currently drive about 3 trillion miles per year and get into about 5.5 million traffic accidents. If I did the math right, driverless cars will result in about 2/3 fewer accidents per year than we experience now. Should we all welcome our autonomous vehicle overlords now?
http://www.usacoverage.com/aut...
http://www.afdc.energy.gov/dat...
I hate these traffic blocking cars with a passion. I live in Mountain view and they are doing 15Mph around a curve which is not even posted as a curve (no danger) and 20-25 where the speed limit is posted at 35-45.
Oh, at first it tolerable. "awe, look at the cute little Google car." After sitting behind these things trying to drive around the city for over a year, I am considering getting an IPO for anti-Google Car chaff cannons. (soda cans full of shredded aluminum foil). I'm pretty sure that would be illegal, so I have not filed a company charter.
-The wise argue that there are few absolutes, the fool argues that there are no probabilities.
Yes but 14 accidents per 2 million miles is basically nothing. If you take truck stats, which is all I could find I could find because it's a nation wide stat, in 2012 there were 77000 crashes with injuries, on 268,000,000 miles, of those 3802 were fatal crashes.
That's NOT an accurate comparison! In fact you cannot compare the two. The google system lets you pick a destination and the car will drive you there without intervention. Many owners who have tested Telsa's system have said you need to pay attention to what it's doing or you could potentially die like that one driver who sadly trusted it enough to watch a movie. The systems are so different on levels of automation it'd be like saying there's more miles on cruise control. Read up and learn the difference.
"15Mph around a curve which is not even posted as a curve (no danger) and 20-25 where the speed limit is posted at 35-45." Do you think they have the speed set low right now so when accidents do occur that statistically there is less risk of a fatality? Can they handle weather, potholes, unmapped roads, and construction yet? I hate the hype, and hopefully soon these cars will drive fast, and work better, but it seems like all of these news reports are always forgetting to mention the bad.
It's sad this was modded down to oblivion but it's tragically very true.
When evaluating who was at fault of an accident it's not who *caused* the accident, but who gets hit in most cases.
If someone's coming into your lane you and you change lanes and get a broadside it's your fault, as opposed to staying straight and praying your car stops in time and getting killed in the head on.
~ People that think they are better than anyone else for any reason are the cause of all the strife in the world.
Yes but 14 accidents per 2 million miles is basically nothing
vs 5.5 million accidents over 3 trillion miles?
https://www.google.com/webhp?sourceid=chrome-instant&ion=1&espv=2&ie=UTF-8#q=how+many+auto+accidents+per+year+in+the+us
http://www.rita.dot.gov/bts/sites/rita.dot.gov.bts/files/publications/national_transportation_statistics/html/table_01_35.html
Nah, call me a Luddite all you want. We're still winning.
And, that 2 million miles has mostly been on the same few miles of carefully mapped roads. Autonomous cars are a great idea, but, they need to demonstrate that they can handle more that the occasional stray pedestrian. They need to demonstrate the ability to drive on unfamiliar roads and roads with unexpected barriers (road construction that wasn't there yesterday) as well as other vehicles that behave in unexpected (and often illegal) ways.
I can't wait to get caught behind a double-wide tractor driving 30mph down the shoulder/lane and have the AI too afraid/unable to pass in the oncoming lane.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
So they automated politicians also?
Table-ized A.I.
12k miles is considered average yearly mileage that's about 32 miles a day so 2 million miles wouldn't be like 300 years of human driving.
It's even worse when you consider that there are over 6 billion humans on the planet and just over a billion cars for them to drive so if the average is 32 miles a day that makes 32 billion miles of human driving every day. So perhaps the headline should read 5.4s of human driving or 167 years of one person driving.
That's why I never use subways, airplanes or trains. If I can't drive, I ain't getting in!
There would be no accidents if we all drove 5 mph, only in good weather, and on mostly straight roads (highways). These on the road are going to mean slower traffic for all.
"...then the traditional taxi services."
Only when Google comes up with a version that speaks Urdu finds the special long way to get to each destination.
I have to admit that I don't buy this. It's comparing normal car crashes reported to police versus google car crashes, which get included in the "14" statistic no matter how minor. Also, you could compare google car crashes with injuries (1 out of 14--too small for meaningful analysis) versus normal crashes with injuries (1 out of 4). Or you could compare the number of crashes caused by humans (almost all) versus those caused by the google car (1).
Don't put them on I-294 or any other IL tollway no one does the 55 or even the 45 work zone
I think that you should only count the 1 crash caused by the autonomous system. That's 1 crash due to a glitch out of 2 million miles.
Unfortunately most of the time people don't swerve to avoid the head on. It's on them so fast they're hit before they can react. I've seen a lot of those accidents on the road I commute to work on. It's a 4 lane highway with a accel/decel lane in the middle. No divider and traffic typically running between 65 and 75 during rush hours. Just a few months ago a man had a heart attack and swerved into oncoming traffic killing another man headed the opposite way. Witnesses said it was so fast that the victim in the other car never even hit his brakes. There have been many of these over the 30 years I've been commuting down that highway. Enough that I ride in the far right lane and don't give a shit how fast I get to work.
And, that 2 million miles has mostly been on the same few miles of carefully mapped roads. Autonomous cars are a great idea, but, they need to demonstrate that they can handle more that the occasional stray pedestrian. They need to demonstrate the ability to drive on unfamiliar roads and roads with unexpected barriers (road construction that wasn't there yesterday) as well as other vehicles that behave in unexpected (and often illegal) ways.
Drive around long enough and you will meet most kinds of crazy, even if you ask a truck driver that's hauled cargo the same route the last ten years. As for the rest, it's not that big a blocker as you might think. Imagine an autonomous taxi with a limited coverage area, it doesn't have to understand the road to your mountain cabin. How about an autonomous shuttle service from the airport to near-by hotels? A hop-on hop-off tourist bus driving the same round over and over? There's a lot of potential even for a limited service, not to mention they could do crowd-sourcing. That is to say, you drive the car where you want it to go, let the sensors collect whatever they need and upload it to Google. With luck they'll process it and next time it'll be an approved destination. I'd take a car that could do my commute, the rest would be nice-to-have but I'd be willing to teach, win-win for me and Google.
Live today, because you never know what tomorrow brings
Yeah sure thing buddy, because we're talking about ALL transportation here, not just automobiles. Like I always get off my bike and walk it down hills, because I'm not actually driving the bike with my legs when I go down hills and that's just WRONG.
Oh, please. What's next? You going to accuse me of never using an elevator or escalator, too? We are discussing automobiles.
300 years of human driving. hmmm that must make my mother just over 200 years old.
Those second million might have happened in my neighborhood in Chandler, AZ. for the past year at least, we've had swarms of them driving around our neighborhood streets at all times of the day and night, thought not late night that I'm aware of (ie mid-5 am, as a linux lead for several large international companies, I get more done at night so I'm up working those hours when no one annoys me with chats etc). At times we've seen 6 driving through our streets all in a line. Driving from pickup up dinner and gas yesterday, about 2 miles, I saw 3 of them including one at the pump next to us at the gas station (we drive a Volt, $13 fill up! W00t!). At first it was interesting, now it's just another car on the road. Haven't heard of any accidents, nor seen any. Though I haven't looked, you might be able to catch some in Google Maps sat images, I see our cars every refresh.
It's very comforting to see them being used so successfully locally, and really gives you some confidence in the product.
Google's cars can cope with the unexpected. A few years ago they showed some detail of how they handle roadworks, and unexpected/illegal behaviour by other drivers and pedestrians has been assumed from day one.
The cars use GPS for navigation and supplementary data, but not for primary driving functions. For that they use lidar and cameras that can read signs and interpret things like cones marking off roadworks (so they don't try to simply drive between cones etc.)
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
This doesn't count all of the times that the human had to take over because the car *would* have caused an accident.