300,000 Miles — Availability by 2017

Google says that its self-driving cars have now gone 300,000 miles with no accidents (except once when one of the cars was rear-ended at a stoplight).

Google released the above video a few months ago in celebration of reaching 200,000 miles. Everything in it seems normal until the car parks in a handicapped parking spot. I thought, “Whoa! Google is going to have to teach its cars not to use those spots.” Then the video revealed that the “driver” was, in fact, “well passed legally blind.” It was a moving demonstration of how self-driving cars will change our lives.

Meanwhile, Ford says that, by 2017, it will sell cars that can drive themselves in certain conditions, specifically heavy traffic. Once just one-in-four cars on the road have this feature, the company says, travel times in traffic will be reduced by 37.5 percent.

Google says it is still figuring out how to deal with snow-covered roads and temporary construction signs. But in normal driving, the main obstacles to self-driving cars is having complete maps of the road network–Google has mapped California and is working on Nevada and Washington DC–and reducing the cost of the “Light Detection and Ranging” (LIDAR) device that sits atop Google’s (and other) self-driving cars. Manufactured by Velodyne, these cost about $75,000.

Mass production should greatly reduce their cost. Velodyne already has a smaller version that costs only $30,000. The $75,000 unit collects 1.3 million data points per second while the $30,000 one collects only 0.8 million per second, but Velodyne says that should be sufficient for self-driving cars.

Share

42 thoughts on “300,000 Miles — Availability by 2017

  1. Andrew

    What is the overall frequency of accidents?

    I’ve driven 450,000 miles in 21 years with two accidents – a skid on black ice on a dark night while going under 50 mph on an Interstate, and being clipped at an intersection by a speeding driver.

    Only around 1:100 people will die on the road after perhaps 1 million lifetime miles of driving so the Google car not having killed anyone yet is not that remarkable. My inner engineer says that several million miles of testing would be needed just to begin to quantify its actual dangers, and whether it is inherently safer or more dangerous than human controlled cars, and at least 500 million miles would be needed to determine if this car is a deadly accident waiting to happen.

    Frank Reply:

    You’re underestimating the factors that cause crashes to begin with. The largest factors are distracted driving and apparently failure to use turn signals, which is an epidemic in WA state and Seattle and a major pet peeve of mine.

    Autonomous cars hold the promise of eliminating these crashes entirely as one, computers don’t get distracted and don’t text or drink lattes, and two, computers will either automatically signal or signaling devices will become obsolete and then signaling devices will truly be optional equipment.

    The amount of accidents caused by real or imagined faults of autonomous technology would logically pale in comparison with the current leading causes.

    C. P. Zilliacus Reply:

    Autonomous cars hold the promise of eliminating these crashes entirely as one, computers don’t get distracted and don’t text or drink lattes, and two, computers will either automatically signal or signaling devices will become obsolete and then signaling devices will truly be optional equipment.

    Presumably these cars don’t consume substances that impair their ability to safely drive on the public highway network either.

    Frank Reply:

    But let me imagine some problems before someone else does. What about terrorist hackers who can hack into the car and make it do crazy stuff like drive itself into a federal building loaded down with explosives?!?!

    C. P. Zilliacus Reply:

    Frank, hacking the software that runs a fleet of driverless cars is definitely a concern.

    Knowing what I know about software generally, were I to get in to a driverless car, I would feel much better knowing that the software was developed by Google and not by Microsoft, the company with operating system products that popularized the GPF and the Blue Screen of Death (BSoD), and provided ways to implement and spread tens of thousands of computer viruses and malware programs.

    Andrew Reply:

    So these cars will never have mechnical or system failures, blow a tire, skid in rain, snow, or ice, misjudge trajectory and position at an intersection or driveway, run over a pedestrian, get in the way of a truck or bus, change lanes into another car also changing lanes, become jammed up at the end of a lane at speed, simply drive off a cliff on a blind curve, etc., etc.? Such blind faith in technology!

    I can’t help but remember watching the first driverless car at Carnegie Mellon struggle with distinguishing a dark barked tree from the asphalt path it was supposed to drive on, forcing the human to repeatedly take command to prevent a wreck.

    Frank Reply:

    Your fearmongering list only distracts away from the primary FACT, which you fail to address or even acknowledge:

    The vast majority of automobile accidents are due to human error.

  2. bennett

    Regardless of the severity, the inevitability of a driverless auto being in an accident where culpability must be assigned brings up several legal and policy issues. There is a hornets nest of political and logistical wrangling awaiting this technology.

    Also, while I really like this idea, I am weary of the claims that driverless cars will solve congestion problems and accidents on the road. These things may be true in a theoretical vacuum but I have my doubts when it come to the real-world application.

  3. davek

    I very much enjoyed that, but it lost some of its punch because I already knew from your post that the driver was blind. Still, it did get me misty-eyed.

    C. P. Zilliacus Reply:

    That is what makes this video so compelling to watch.

  4. C. P. Zilliacus

    Just one personal grip about this (and presumably other) self-driving vehicles.

    They all (presumably) have automatic transmissions.

    I say that as a lifelong driver of vehicles with clutches and standard transmissions.

  5. C. P. Zilliacus

    Speaking of self-driving cars, it has become an election issue in one race for state office in Florida, according to this article in Salon.

    Frank Reply:

    Didn’t watch the video as I have no sound on this eight-year-old Dell I just refurbished (sound is next–maybe). But what I gleaned from the article is that it is all about “hints” of danger and fear. Fear. FEAR. FEAR! After all, that’s what Amurika is all about!

    C. P. Zilliacus Reply:

    Frank, I agree with you more than I disagree with you regarding the fear factor.

  6. LazyReader

    I agree, a million miles worth of testing. When Ford unleashed EcoBoost for it’s cars, they put a million miles equivalent testing on it before trying to put the engines in their cars. That’s just engines. A driverless car will be made of subsystems, all of which must be tested and since software isn’t something you can feel it’ll be the most painstaking thing to test. How often do we complain of software glitches. It’s like those people who have a feel of the road or listen to their cars thinking “time to change the oil, tires or get it tuned” just by listening.

  7. Sandy Teal

    What does the Google car do when street hooligans pretend to wash the windshield and demand payment? If it has a force field, or electrified body armor then I totally buy into this car.

    C. P. Zilliacus Reply:

    ;-)

    LazyReader Reply:

    How did KITT get rid of panhandlers.

  8. Dave Brough

    Combine it with elevated guideway http://faculty.washington.edu/jbs/itrans/third%20generation.htm
    and In one fell swoop you would have had automobile-like transportation with:
    Universal access
    Automated Safety
    Small, inexpensive cars
    Freight delivery / refuse collection
    10 times the mpg
    Electric power
    Support for high density cities
    50 times less need for road area
    10 times greater parking density
    Intercity travel at 150 mph (no need for HS rail)
    Private investment in transportation
    The best part? Transit is obsolete.

    Andrew Reply:

    The cost of building elevated guideway in urban areas is over $10,000 per foot. Who in the world is going to pay for that and who wants it located outside their home?

    Dave Brough Reply:

    $10,000 a foot, Andrew? Where did you get that figure? My sources tell me $2.5M/mile. Compare that to the $70M/lane-mile for E-way and $100M/mile for rail.
    That’s not figuring the cost of death/mayhem into the cost of existing road/transit, either.
    As for “who would want it located outside their home?”, no one. Robocar drives you the few miles to the closest guideway, which would be located alongside existing E-ways, along floodways (LA), and otherwise out of sight as much as possible.

    Andrew Reply:

    It costs $2.5 million per mile to just pave a road with asphalt!

    Structures as shown in your picture would cost hundreds of thousands for each bent. Which is to say nothing of them being absurdly slim for the loads they will carry. You would need to pay big bucks for utility relocation and the electronic systems needed to operate, power, and govern the system.

    Frank Reply:

    Andrew, why do you insist in inventing statistics? You need to start citing your sources and providing evidence rather than making up stuff.

    A quick search of Google revealed these tidbits:

    “Generally speaking, asphalt paving costs approximately $280,000 per mile, but can vary widely depending on the current asphalt price, road width and asphalt thickness.”

    “An asphalt overlay can range anywhere from $95,000 to $130,000 per mile depending on the width of the road and cost of the asphalt.”

    It’s a bit more in Sacramento, of course: “For well-maintained roads, the pavement rehabilitation cycle, meaning an asphalt overlay, comes due in 15 years for arterials and 30 years for local streets, costing $300,000-$400,000 per mile.”

    You’re way off with this $2.5M figure.

    Andrew Reply:

    Frank:

    Pavement overlay is a 2-3″ milling and replacement of the top wearing surface.

    To build a road full depth requires an dense compacted aggregate base and over one foot of total asphalt thickness in base coat and wearing coat.

    One recent bid I received for work was $100 per sq. yard to install pavement on an already graded subase. That is $1.75 million per mile for a 30 ft. wide roadway. Add in aggregate and grading, and you get the price I previously noted.

    Frank Reply:

    “It costs $2.5 million per mile to just pave a road with asphalt!”

    Can’t re-write history. So what you were REALLY saying, and making your reader infer, is that it costs $2.5 million per mile not “just to pave” but to build a road.

    Word choice matters.

  9. Dave Brough

    Too bad Google didn’t have an airliner application a couple of years back when, after a certain distracted pilot flew through a flock of geese and then through a series of comically-wrong – but fortunately not tragically-wrong – moves, turned a perfectly good aircraft into a submarine.

  10. Dan

    after a certain distracted pilot flew through a flock of geese and then through a series of comically-wrong – but fortunately not tragically-wrong – moves

    Here on planet earth in this reality, that never happened.

    And I LOLzed at the comical elevated guideway too. Some planets must have many more resources than ours!

    DS

    Dave Brough Reply:

    OK, Dan, What DID happen? On my planet, both pilots were gawking “the beautiful view of the Hudson” (transcript: Sullenberger). Instead of returning to LGA – perfectly do-able, according to the NTSB (and every pilot that sim’d the event) – they splashed. What did I miss?
    And what’s so ‘commical’ about elevated guideway?
    DB

    Dan Reply:

    First, I’d like to know how you’d take evasive action in an Airbus to avoid the geese you might have seen. That would be very interesting how you’d guide that plane in just a few seconds to avoid a bird strike. Very interesting indeed!

    Second, I’d like to see your evidence from the NTSB that returning was doable, and that the NTSB found the outcome could have been avoided.

    And on what planet are the resources for an elevated guideway as outlined above? This one? Srsly?

    DS

    Dave Brough Reply:

    As for the geese being “seen”, it was mid-afternoon and conditions were CAVU (Ceilings and Visibility Unlimited). A dark, flapping target 2m wide and 100m long and silhouetted against the blue sky makes for a pretty easy target – provided, of course, you’re looking.

    As for “taking evasive action in an Airbus”, those aircraft are pretty nimble. Just ask Bob Hoover http://www.youtube.com/watch?v=Xp2Uc9XvmjY or someone’s that’s actually rolled a big jet, Tex Johnson http://www.youtube.com/watch?v=Ra_khhzuFlE

    As for ‘How (to avoid)’, that’s easy. First, instead of gawking the view down (of the river, as noted) you use the set of MK-1 eyeballs God gave you to scan the view ahead. See and Avoid are the two guiding principles of aviation.
    Second, you always have a constantly unfolding Plan B. Sully had nuthin’

    As for evidence from the NTSB that returning was doable, this from CBS News. NTSB: Sully Could Have Made it Back to LaGuardia
    http://www.cbsnews.com/2100-201_162-6459709.html

    As for providing “Evidence that the NTSB found the outcome could have been avoided”. Nothing from the NTSB, just my own deduction. Based upon the following:
    The transcript shows that Sullenberger only saw the birds one second before impact. Extrapolating the 220 kts they were traveling works out to about a football field. Were they looking, they could have picked the birds up handily at 3 miles, this translating into 45 seconds. Even 15 seconds (one mile) would have given an alert pilot sufficient time to nudge out of the way. The pax wouldn’t have even noticed.

    All of this brings us back to the Antiplanner’s posting: Robocars and safety. From the day after we get our commercial airliners under robotic control, aviation incidents will decrease by 80%.

    As for “On what planet are the resources for an elevated guideway as outlined above?”. Steel (for rebar and rails) and cement are both cheap and plentiful and available on this planet. We’re talking about a structure engineered to carry lightweight vehicles on a smooth guideway, not 80,000 pound trucks thumping down a road. Big difference. Srsly.

    Dan Reply:

    First, the NTSB found the plane could not have returned in their official report.

    Second, you are one amazing dude to be able to pick out a flock of geese at three miles against an urban background in your scan. Superman, surely. If there were more pilots like you, we’d never have bird strikes, surely.

    Thank you ever so much for sharing your concerns and your transport solutions!!!

    DS

    Dave Brough Reply:

    Dan.
    “the NTSB found the plane could not have returned in their official report.”
    Read the report, Dan. It was only after a 35-second delay that a return was unsuccessful. When the turn was made promptly, ALL sims returned safe and sound.

    “you are one amazing dude to be able to pick out a flock of geese at three miles against an urban background in your scan. Superman, surely.”

    There was no ‘urban background’. They were below and heading upwards on a northwest heading. This was an unusually large flock at 12 o’clock and high and sticking out against a clean background like a sore thumb. Easy pickins, and yes, at 3 miles – provided, of course, you’re looking. Those guys weren’t.

    “If there were more pilots like you, we’d never have bird strikes, surely.”
    Sarcasm aside, Dan, most bird strikes are small birds at low altitudes – typically on liftoff – and are not preventable. This particular strike occurred at altitude and involved large birds flying in a known migratory flyway. It was preventable.

  11. Frank

    On August 15th, 2012 at 8:31, Andrew said:
    “Such blind faith in technology! I can’t help but remember watching the first driverless car at Carnegie Mellon struggle with distinguishing a dark barked tree from the asphalt path it was supposed to drive on, forcing the human to repeatedly take command to prevent a wreck.
    On January 30th, 2012, Andrew said:
    The limitless faith in technology being expressed here is astonishing. I used to watch the early version of these systems being tested by Carnegie Mellon engineers in Schenley Park in Pittsburgh. The system had trouble distinguishing a dark tree from the asphalt path it was supposed to follow. They later took the vehicle out on the highways at night, and eventually took a cross country trip and proudly announced it had self-driven 98% of the way. Wondeful. Watch out if you were around during the other 2%.

    You’re repeating yourself, Andrew. Or should I say The Highwayman? Is he your sockpuppet? He used to sign his posts as “Andrew”…hmmm…would explain the repetition of essentially the same points over and over.

    Speaking of Carnegie Mellon. So, you were there, huh? Have pictures to share?

    BTW, was it an “early version” or the “first driverless car”? Alvin was tested years before Navlab 1, and the Japanese tested a vehicle a decade before Navlab 1. Also Navlab cars were semi-autonomous, not “driverless”.

    Personally, I think you’re pulling this $h!t out of your @$$. We get you’re afraid of technology; no need to guffaw anymore at our “blind faith” (which you’ve projected upon us from your place of fear). Only a scared person would attack primitive 1980s technology thinking it’ll somehow refute what 21st century technology can achieve.

    Andrew Reply:

    Sure Frank. Whatever. Does it even matter what I tell you? Its not like you would believe it anyway. You must be pretty insecure to need to attack me instead of discussing the topic.

    Frank Reply:

    If you equate cliches and logical fallacies with discussion… I’m not attacking you, just your repetitive fear-based posts ranting against technology and demeaning its proponents as blind (figuratively–although, in this video literally) and religious nuts. Great discussion, Andrew!

    Andrew Reply:

    The system combines information gathered from Google Street View[citation needed] with artificial intelligence software that combines input from video cameras inside the car, a LIDAR sensor on top of the vehicle, radar sensors on the front of the vehicle and a position sensor attached to one of the rear wheels that helps locate the car’s position on the map. As of 2010, Google has tested several vehicles equipped with the system, driving 1,609 kilometres (1,000 mi) without any human intervention, in addition to 225,308 kilometres (140,000 mi) with occasional human intervention. Google expects that the increased accuracy of its automated driving system could help reduce the number of traffic-related injuries and deaths, while using energy and space on roadways more efficiently.[1]

    http://en.wikipedia.org/wiki/Google_driverless_car

    Note the ratio of autonomous driving to “mostly” autonomous driving, something all of you boosters not only fail to mention, but actively deny occurring.

    Frank Reply:

    “Note the ratio of autonomous driving to ‘mostly’ autonomous driving, something all of you boosters not only fail to mention, but actively deny occurring.”

    Please show that a proponent of this technology has in these comments denied that the current technology is semi-autonomous.

    And my research extends beyond Wiki to the actual 1987 Carnegie Mellon report showing the semi-autonomous technology used in Navlab 1 can recognize trees:

    Section III, Introduction: “The system built from these tools is not capable of driving the Navlab along narrow asphalt paths near campus while avoiding trees and pausing for joggers that get in its way.”

    Figure 17: “The navigable area is shown as a mesh, and the two trees are detected as ‘textured obstacles’ and shown as black polygons.”

    So nothing from the 1987 report about Navlab 1 being unable to distinguish “a dark barked tree from the asphalt path it was supposed to drive on.”

    It really doesn’t matter if you’re making this up, anyway, as current technology has advanced far beyond the 286 processors (with only 512K of local RAM and paltry 1 MB of system memory) used in Navlab 1. And it will continue to progress. And someday, we may have fully autonomous cars. Despite all the fear mongers who compare 2012 computers and technology to those used in 1987.

    Andrew Reply:

    Frank:

    Please show that a proponent of this technology has in these comments denied that the current technology is semi-autonomous.

    Every discussion and article of this technology ends up along the lines of the driver not actually having to drive, but being able to kick back and text or read or watch a movie because the car is driving itself without human intervention. As opposed to reality, which is 1% autonomy, and 99% needing active human monitoring.

    And I don’t need to “research” something I observed with my own two eyes and discussed with people involved.

    Frank Reply:

    “Every discussion” is not evidence; it’s a generality.

    Again, please show using evidence in the form of quotations that a proponent of this technology has in these comments denied that the current technology is semi-autonomous.

    When speaking of “being able to kick back and text or read or watch a movie,” it’s a reference to the POTENTIAL of this technology.

    “…I observed with my own two eyes and discussed with people involved.”

    Ah, anecdotal evidence. Do you have any documented evidence? Again, it really doesn’t matter because it’s an attempt to discredit 2012 technology by finding fault with 1987 technology. That is: it’s a fallacy.

Leave a Reply