Self-driving cars


I tend to be swayed back and forth by the promise of self-driving cars. On the one hand, I read about how good they are and have advanced so much that one might expect them to be available for commercial use within the next decade. Then I read that that they are only as good as the latest map updates and cannot cope with the kinds of temporary changes in road conditions that are common and then I feel pessimistic that they will be a reality soon.

But now there is a report of a car that took a 9-day trip from San Francisco to New York City, a 3,400 mile journey where it was driverless for 99% of the way, although it always had a driver in it at all times. The car was designed by a British company named Delphi.

To make a 2014 Audi SQ5 fully automated, engineers installed six long-range radars, four short-range radars, three vision-based cameras, six laser range detectors and a full suite of software. That’s a lot of technology, but one would be hard-pressed to spot it, as it all blends seamlessly into the vehicle’s design.

Throughout the drive, Delphi’s vehicle encountered weather, construction zones, bridges, tunnels and, of course, reckless drivers. However, the vehicle handled these situations better than expected, and engineers rarely took control of the wheel.

While this looks good, the catch is of course the remaining 1% or how to handle the situations when the engineers had to take control of the wheel. From other reports, it appeared that this car was automated only on the highways and human drivers took over when the car took an off-ramp into local streets. Whether one is optimistic or pessimistic about the future of self-driving cars depends on how difficult that remaining 1% turns out to be.

Here’s a short clip of what it is like to be in such a car.

The car drives in such a steady way, with minimal changes in speed and direction, that it seems actually boring, behaving pretty much like it would with a very law-abiding driver.

The car, according to Delphi, never broke a speed limit, which apparently did not go over well with other drivers during the trek. Owens acknowledges the vehicle was the recipient of a “few hateful gestures.”

Delphi put a human driver in the driver’s seat during the ride given the test aspect and to work out a few kinks, such as the car not wanting to move into a crowded left lane to avoid a police stop on a road shoulder and edging to one side for some looming semis, as apparently the vehicle shares the same fear of huge tractor-trailer trucks that many humans have.

Comments

  1. sundoga says

    These things will only work if they’re mandatory. We’ve gotten to the point of being able to deal with pretty much any road conditions – what they can’t. and probably will never be able to deal with is the average driver, who thinks he’s good at it but is actually abysmal.

  2. brucegee1962 says

    The real problem comes with liability. Even if they’re ten times better drivers than we are, they’ll never be perfect — a few accidents will be unavoidable. Who pays — the software designers, the government, or what?

  3. says

    Sorry, but for me, the idea of self-driving cars is perhaps one of the most exciting innovations on the way. I’m fully aware of the hurdles that still need to be jumped, and technology is technology. But the whole idea is just damn perfect. It sometimes amazes me that there are people who are actually against the idea when, really, it should be common sense by now. You need to drive from California to New York?

    No problem! You can even sleep… don’t worry, the car will let you know if you need to take over. No more wasting money on motels… you can drive straight through (well… mostly… you’re gonna want to eat, and if these aren’t electric cars, fuel up [although, really, they should be fully electric cars]).

    Tesla Motors is already innovating this. Their Model S basically has auto-pilot:

    Autopilot combines a forward looking camera, radar, and 360 degree sonar sensors with real time traffic updates to automatically drive Model S on the open road and in dense stop and go traffic. Changing lanes becomes as simple as a tap of the turn signal. When you arrive at your destination, Model S will both detect a parking spot and automatically park itself. Standard equipment safety features are constantly monitoring stop signs, traffic signals and pedestrians, as well as for unintentional lane changes.
    Autopilot features are progressively enabled over time with software updates. The current software version is 6.2, adding automatic emergency braking and blind spot warning.

    I want a Tesla so bad it’s not even funny… but I can’t even come close to affording it… 🙁

  4. lorn says

    They really need to nail that other 1%. A vehicle that does fine 99% of the time, just to lull you into complacency, and then, at an entirely random interval, tries to kill you isn’t going to get it for most people.

  5. chigau (違う) says

    Doing fine 99% of the time and trying to kill you the other 1% sounds like human drivers.

  6. Just an Organic Regular Expression says

    Inspired by some discussion of this on a recent SGU I’ve been thinking about some of the possibilities. Some trivial points that I think are entertaining to ponder.

    First: Presuming people will still own personal cars at all, it will reduce the number of cars owned from its present 2.2 per family in the US. Think about it: people have more than one car because family members go in different directions. But there’s no reason for your commute vehicle to sit idle in a garage all day waiting for you to finish work. What a waste! Your one family car can drop you off at work, spin around and zip home empty, to be available for use by others all day.

    Second, that family car doesn’t have to be dedicated to your family all day; if you don’t need it, it could automatically log in to Uber or Lyft and earn money driving other people around for hire, automatically, before coming to pick you up at the end of the day. And at night when you are sleeping, your car could be out earning its keep!

    Third: when the infrastructure adjusts to self-driving cars, it will be the end of parking problems! You want to eat at a nice café in downtown SF, or in some village with narrow streets? Your car drives you to the curb at the site; you hop out; the car, without further instruction from you, toddles off following beacons to the municipal garage, which might be kilometers away, and parks itself. As you wait for the check, you pull out your phone and call your car, which un-parks itself and comes around to pick you up — or picks you up at the end of the street so you can have a nice stroll after your meal.

    Fourth, a big problem I have not seen addressed is the “last 100 yards”, when your destination is up a driveway, behind a building, in an alley or otherwise off the mapped roadway. Recently I attended an event at a big arena: parking was in a vacant lot nearby, with attendants waving the stream of arriving vehicles into parking spaces across a large expanse of grass and gravel, without striping or other guidance. A true self-drive car like the one Google has shown has no steering wheel at all. How can it be guided in this situation? Or recently, a truck caught fire on a local bridge, and police rerouted freeway traffic off one ramp and back up the next. How to communicate such short-term, un-mapped, ad-hoc route changes to the computer in the car?

    It seems to me one answer is a standard, preferably an international standard, on coded signs and batons. For example the car’s computer should be able to detect a specially-striped baton and follow its instructions just as airline pilots follow the instructions of the da-glo batons used by the ground personnel at the airport. A black and orange “Detour” sign with an arrow might be enough but it could be augmented with either special striping (bar codes?) or wireless signalling.

    But that raises the awkward question, if a car will automatically follow such signals, what’s to keep pranksters or criminals from misleading trusting cars?

  7. md says

    Goodbye long haul 18 wheeler truck driving jobs. That was a nice paying gig for non-college types.

  8. TGAP Dad says

    As a Michigander, self-drive seems a ways off to me. From the icy, snow-covered roads in winter to the badly faded lane markings to our craters-of-the-moon potholes, I see a lot of situations a self-drive algorithm would have to master before I’d be convinced.

  9. brucegee1962 says

    @7: As someone on the downhill side of fifty, the biggest deal for me is this one: what is the #1 fear of seniors? Losing mobility, and thus, independence. With self-driving cars, people will be able to stay in their own homes a whole lot longer.

    This will also be a big boon for online shopping. Just click on the groceries you want delivered, and they’ll show up at your door.

  10. Mano Singham says

    brucegee1962,

    It may also reduce the need for extra cars for a single family since people could take the car somewhere and then send it back without a driver if someone else needed it.

  11. Just an Organic Regular Expression says

    @10: I expect long-haul trucking to be the first kind of driving to be automated. The routes are mostly well-mapped highways. The economic benefits are huge, for example automated trucks can run 24/7 whereas human drivers have a legal requirement for a certain number of hours of sleep. Automated trucks can coordinate to make nose-to-tail “trains” that have fuel efficiencies from reduced air resistance. However, I also expect that there will be a patchwork of laws such that for a long time after they are really needed, at least some states will require a human — awake, and not reading or playing games — back of the wheel, and this will cut into the savings. Also, the “last hundred yards” problem is especially acute for trucks, because they have to back up to loading docks in restricted alley spaces.

Leave a Reply

Your email address will not be published. Required fields are marked *