Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Transportation Google

How Tesla's Autopilot and Google's Car Are Entirely Different Animals (robohub.org) 142

Hallie Siegel writes: Developers and futurologists have long talked of two paths to autonomous cars: the incremental path (where autonomous features such as adaptive cruise control, autonomous parking etc are slowly added to make the car increasingly autonomous) and the revolutionary path that abandons the human driver altogether — the Google car approach. Robocar expert Brad Templeton compares Tesla's latest autopilot technology to the approach Google is taking, explaining why some people think autonomous cars are still decades away, while others believe they are just around the corner.
This discussion has been archived. No new comments can be posted.

How Tesla's Autopilot and Google's Car Are Entirely Different Animals

Comments Filter:
  • by turkeydance ( 1266624 ) on Thursday October 29, 2015 @07:28AM (#50824003)
    yet?
    • by Anonymous Coward

      One is, the other, according the TFS, is an entirely different animal.

  • Very different (Score:5, Insightful)

    by Viol8 ( 599362 ) on Thursday October 29, 2015 @07:37AM (#50824047) Homepage

    One looks like it was designed for adults, the other for toddlers. The google car couldn't look any more childish if had pedals inside and coloured wheels.

    • It's a good thing it's not a fashion contest.

    • by Anonymous Coward on Thursday October 29, 2015 @07:52AM (#50824141)

      One looks like it was designed for adults, the other for toddlers. The google car couldn't look any more childish if had pedals inside and coloured wheels.

      The Google car is designed for American adults. The look like Weebles and act like toddlers.

    • Yep because the shape is the major technical challenge to autonomous vehicles.....

      • by Viol8 ( 599362 )

        If you want anyone to buy the thing then you'd damn well better get the styling right!

        • Re:Very different (Score:5, Insightful)

          by leonardluen ( 211265 ) on Thursday October 29, 2015 @09:13AM (#50824737)

          who cares what googles car looks like now. when google gets the sensors ans software working then someone that has a sense of style will come along and design one that looks good.

        • If you want anyone to buy the thing then you'd damn well better get the styling right!

          If you want anyone to buy the thing you need to consider putting it up for sale in the first place. Is the idea of a concept car lost on you? I mean every car company produces cars like this. No one actually ever puts them on the market.

          So chillout, I'm sure when the technology is ready you'll get something that looks normal.

    • That would be because the Google car isn't designed to be a commercial product- it's designed to introduce fully autonomous cars to the public consciousness in as non-threatening a manner as possible. Sure, most people probably wouldn't want to actually buy a car that looked like that, but they are for more likely to be sold on the technology in the form of a cute, easily anthropomorphic demo vehicle, and then look to see if there's a a "stylish" model available.

      Furthermore, they seem to be heavily targeti

  • Just this morning, after it's not been driven for about six years for various reasons, I paid a very large garage bill for fixing up my 1991 Honda Civic.

    This car has no engine ECU, no ABS, no airbags, no lane assist, no automatic braking, no shit at all. What it DOES have is four wheels, brakes, lights and something to steer it with. It also has twin carburettors and a manual choke.

    First job was to fill it with petrol, and as the engine warmed up I started to remember just how good this old car is to drive.

    • by delt0r ( 999393 )
      good luck fixing a carburetor.Or the fuel pump. How much oil does it leak? Thing about the good old days is, they weren't.
      • Right, they weren't old at the time, but they are now.
        • by delt0r ( 999393 )
          Even when a carburetor is not old. They are very complicated and difficult things to fix. I know, i did several times on my 1974 Hilman hunter. The later ones where even worse. My 1990 motor cycle carb had more parts, most moving than you could shake a stick at. Fuel inject is a far simpler system.
    • Weird, my US-spec 1987 Civic Wagovan had FI... I take it Honda continued selling carbureted engines elsewhere for a while or something?

  • by Anonymous Coward

    I hope they're just around the corner, but they'll be out of reach of most people. Like Tesla now.

    Maybe by the time I retire they'll be reasonable so I can tour the country and never even drive.

    • combine this story about wireless EV charging [slashdot.org] autonomous cars and a crazy software developer and i wonder how long it will be before someone programs their autonomous car to drive them around in an infinite loop after they die...or at least until the car breaks down.

      • by Anonymous Coward
        piano wire around the neck, through the window, and tied to a pole, program the damn thing to take you to your bitch ex's house and let her see just how bad she hurt you
  • by sinij ( 911942 ) on Thursday October 29, 2015 @07:57AM (#50824165)
    Google approach of map everything in excruciating detail has one big flaw - it assumes that we could know in advance how the driving environment would look like and navigate based on this. This is not a reasonable assumption, simply because mapping every back road is gargantuan task. Therefore, we will end up with 'good enough' Google cars, that can drive major roads and urban centers.

    Tesla (and all other car manufacturers) approach is to have car react to environment with little advanced knowledge. This is gargantuan task and is still computationally unfeasible. Therefor, we will end up with 'good enough' Tesla-like car, that can drive anywhere but still require driver's supervision.
    • by crow ( 16139 )

      Human driving is a mix of both methods. When you're on a street you're familiar with, habit takes over, and you barely notice what you're doing. On an unfamiliar street, you're much more active as a driver. At some level, humans require driving situations to be predefined, in that they need to match a familiar template. Road designs are all standardized.

      In other words, the more information you have about the driving conditions, the simpler the problem. If you have a map, then you need to watch for anyt

    • Google approach of map everything in excruciating detail has one big flaw

      That would be a big flaw... if it were Google's approach. It's not. Google's system analyzes what it sees in real-time. At present Google is sticking to areas that it has well-mapped, but that's not essential for the vehicle to operate.

      • by sinij ( 911942 )
        Maybe you know something I don't?

        My understanding is that presently Google cars won't self drive outside of specially mapped areas. These are not street-view maps, but detailed telemetry scans that have to be frequently updated to stay relevant.
        • by Cyberax ( 705495 )
          They don't do that _right_ _now_, but that the goal.
        • My understanding is that presently Google cars won't self drive outside of specially mapped areas.

          Presently, this is true. But it's not crucial to the system design. It's not part of the approach, just part of the current testing plan.

          Also, as another commenter pointed out, mapping the whole world isn't as unreasonable as all that. Cars would still have to be able to deal with it when stuff changes (as they do now; even in the limited test areas changes happen faster than updates, and the cars do fine anyway), but that could be the exception-handling case, in which case the car would simply become a l

  • by DickBreath ( 207180 ) on Thursday October 29, 2015 @08:08AM (#50824225) Homepage
    . . . where the speaker compared the two approaches like this:

    Gradually trying to move towards driverless cars instead of working directly on that goal is like thinking that by practicing jumping and getting better and better a jumping that you'll eventually be able to fly.
    • by Richard Kirk ( 535523 ) on Thursday October 29, 2015 @08:22AM (#50824315)

      Nice.

      Google and Tesla are doing different things for good reasons. Tesla makes electric cars, and it needs to go carefully or it will lose its core business and customers. So they start from an electric performance car and gradually work up to an autonomous performance car. Google doesn't make cars, so it is not risking a core business; and their potential customers are mostly people who can't drive or don't trust their eyesight any longer, so anything that lets them potter to the shops is better than nothing. So they start from a new antonomous car, and work up to an autonomous performance car that can play chicken with the Audis on the autobahn.

      Two different approaches. One of them is not necessarily wrong.

    • by orasio ( 188021 )

      Gradually trying to move towards driverless cars instead of working directly on that goal is like thinking that by practicing jumping and getting better and better a jumping that you'll eventually be able to fly.

      Worked for Superman, didn't it?

    • I would characterise it differently. One is trying to engineer out all the risk, while the other is going to shape the perception of the risk in the market.

      The reality is that no matter how long Google spends trying to make their cars safer and safer (and apparently it is already significantly safer than a human) one day their car is going to have a serious accident. Maybe it is not even the car's fault, and some grandma has a heart attack and smashes into the side of it and dies. It doesn't matter, at that

      • Car crashes kill millions of people a year. Airlines full of passengers go down in horrible, flaming wrecks. It will make news, but I think you understimate how many people are willing to use an extremely useful technology even if there's a slight risk of death. Look at what happens when a serious car defect is found, like shrapnel in airbags, or brakes that don't work... a big fuss is made, sure, but no one stops driving their cars, perhaps except for people with the specific models affected.

        Keep in min

    • From my experience with managers/executives who talk like that, they have little practical experience and will be gone shortly after their first spectacular failure. Slow and steady wins the race. You see some interesting counter-examples from startups, but in general just incorporating the latest technology into existing systems is the way to go.

      • by sinij ( 911942 )
        When you factor-in the sheer number of failed startups and come up with success:failure ratio, I don't see how you can call it a "counter-example".
        • I was just anticipating the inevitable responses along the lines of, "Oh, yeah, what about xxx?" As you say, there are far more failures than successes, and no one is claiming that the telegraph would have been invented by adding electronics to a horse :)

    • <quote>. . . where the speaker compared the two approaches like this:

      Gradually trying to move towards driverless cars instead of working directly on that goal is like thinking that by practicing jumping and getting better and better a jumping that you'll eventually be able to fly.</quote>

      Which isn't incorrect with incremental development.
      If I'm able to increase my jumping height, eventually I'll be high enough to enter orbit, and I'm easily able to fly.
  • by monkeyxpress ( 4016725 ) on Thursday October 29, 2015 @08:09AM (#50824227)

    Driverless cars are going to be hugely disruptive to a large number of industry. The first Uber/Lyft like company to get them going will be able to undercut every taxi service in the entire developed world. We are talking a billion dollar industry there. They will likely be able to gain a significant share of the existing public transit spend in almost every city in the world - even those with good public transit infrastructure - another billion dollar industry. For many individuals it will be more attractive to spend the considerable amounts of money they currently spend owning a private car on an automated taxi service, which is another billion dollar industry.

    The first trucking company to use driverless cars will be able to run trucks more often, for cheaper, undercutting everyone else. This is again a billion dollar industry. Eventually companies like Amazon and Walmart will have vending machine vans that circulate around an area and come to your door with milk and bread faster than you can walk down to your local store. This will change the nature of bricks and mortar retail again. That is another billion dollar industry.

    A fleet of driverless taxi services would potentially make the economics of electric cars look unbeatable. The high load factor of taxis means that you can afford to pay a lot more in capital costs in exchange for massively reduced operating costs. Automated taxis could also manage their own charging, and with apps that pre-plan journeys the car sent to you would be able to ensure it had enough charge to get you to your destination, eliminating the main problem with electric cars right now. The system could probably be built with small (cheap) onboard batteries and a limited number of swapping stations throughout a city. This could massively undercut both gas cars and private vehicle ownership without any further reductions in battery prices. Now you are talking about a trillion dollar industry.

    There is no doubt that driverless cars are a challenging technology to develop. It will be extremely difficult to make them as reliable and safe as we would all like them to be. But in the end, as long as they are, say, an order of magnitude safer than human drivers, the massive economic benefits (i.e. potential profits) will ensure that they are put on the roads. When there is this level of money to be made, capitalism will find a way.

    • by noldrin ( 635339 )

      Other industries will try to push back against them such as municipalities who earn income through traffic enforcement, and insurance companies who make big money on point surcharges.

      Self driving cars will likely change our relationship with our cars as much as media on demand has with our TV. When we drive somewhere in the city, the car drops as off, looks for parking on it own, and can just keep driving around if nothing is found. Heck, if are going for the day, why pay $35 when it can drive to a further

      • Actually, when self driving cars become reality, the last thing you will want is to actually own one. The biggest problem I see is in that the cars will be able to be remotely disabled. The privacy and software security implications are ginormous. Deviatons from privacy and the hacking concerns will make all the current hacker's issues look like spilt milk.
    • by tchuladdiass ( 174342 ) on Thursday October 29, 2015 @09:30AM (#50824865) Homepage

      Not to mention, that self driving cars will very rarely commit traffic violations (speeding, etc). That will dry up a major revenue source for a lot of smaller towns, another billion dollar industry.

      • by delt0r ( 999393 )
        people say this a lot. Got any data on that. And citation if you will.
        • people say this a lot. Got any data on that. And citation if you will.

          Just for speeding tickets: 6 billion dollars: http://www.statisticbrain.com/... [statisticbrain.com] An average of $152 seems a bit low to me.

          If you look at parking meters in San Francisco http://priceonomics.com/san-fr... [priceonomics.com] you'll see they get about $50 million for paid parking, and get $80 Million in parking violations per year.

          • by delt0r ( 999393 )
            The speeding ticket link is not loading. I assume that is for the whole US. And as for San Francisco, 80M sounds like a lot, but for a city of that size it a line item. 6B for a nation (~300B month just for the Fed + local government) is so small you may as well worry about the lack of tax from buggy whip manufactures.

            Sure it a lot of money on its own. But in the bigger scheme of things it just isn't.
      • by dave420 ( 699308 )
        Which is a reason more advanced societies don't allow for such a ridiculous conflict of interests.
      • Not to mention, that self driving cars will very rarely commit traffic violations (speeding, etc). That will dry up a major revenue source for a lot of smaller towns, another billion dollar industry.

        Actually, at first, you'll see the reverse:
        Oh, that car in the next lane is driverless - so you know it's safe to cut it off. Stopping at a stop sign and the car to your right is driverless: you know it's safe to go ahead of it. etc.. Driverless cars will start uploading driving violation videos to you tube or some police agency, cities wanting their cut, will start writing tickets as fast as possible based on these videos. Very quickly, drivers will start behaving - but road death rates for the human dri

    • by Kjella ( 173770 )

      Flying cars would also be disruptive, it doesn't mean it'll happen any time soon. Maybe it'll get approval to run on a few well painted, well lit, pre-approved roads in sunny California during daytime genuinely by itself, but from there to an all-road, all-condition vehicle that people can actually use as a full substitute for driving themselves will take many, many years. It's a long step from a test drive with humans that can take over to saying we're confident this can drive by itself, take the back seat

      • What would be wrong with restricting a driverless taxi service to only certain roadways that have been configured to support them (if required)? You could just start with suitable inner city CBD areas, and build out the capacity with a combination of roadway improvements and upgrades to the car so that it can service larger and larger areas. Similarly, the first automated truck systems could just travel between terminals at the ends of motorways (which are already well formed). Human truck drivers could pic

    • capitalism will find a way.

      Maybe but capitalists don't solve technical problems; engineers do.

      As as aside, the writer states that because Tesla's tech is evolutionary in nature, and this "problem" clearly requires a revolutionary approach, that means it's going to happen quickly. To use [what's bound to be a terrible analogy], just because it'd take forever to come up with a neon-green German Shepherd through selective breeding does *not* mean we'll be seeing genetically-engineered custom-colored dogs on the market within the next f

    • I think this will come eventually. I suspect the solution will involve lots of radar, forwards and backwards, along with several pairs of 3d cameras. The problem with lidar is that it doesn't work in snow or heavy rain. Humans drive based on 3d vision, so why not have computers do it. Problems could arise if the cameras get blocked, but the same think can happen to human drivers. If there are multiple pairs of cameras, perhaps with mechanisms to clear their own view to the outside, they could be quite

    • And these JohnnyCabs won't have to be able to go everywhere.
      If they can just navigate between the airport and all the hotels in the city, that alone would be able to kill the taxi business.
  • Tesla pushed the "autonomous driving" out in a beta update. Immediately youtube was filled with videos of Tesla's with auto-driving enabled almost crashing as it can't handle corners, swerved into other lanes etc.

    Then you have google who is taking a multi year approach of refining the technology before even letting consumers see it.

    Tesla is (in my mind) looking hard to make noise about their products and it is dangerous to push out alpha "self driving" software out to the masses. It is quite a difficu
    • None of the videos that I can see fail when used on the highway, which is the only place Tesla Autopilot is supposed to be used.

  • I use a car mostly for going on the road, but sometimes, it needs to drive other places. I go to events where we park in the grass. I go down driveways that are a mile long. I have experienced emergencies where I have to go backwards down the highway. I just don't understand how completely driverless cars will work in these cases. I have driven a car with adaptive cruise control and emergency breaking. It was fantastic. Please continue down this road and make these features required.
  • When discussing cars, slashdot needs to compare them as animals?
  • Seriously, stop listening to these cretins. They give you fairy-tales and other stupid stuff they believe in. They do not give anything of any worth, but drown out people with an actual clue of what is possible and what not and how long things can be expected to take.

  • They're entirely different kinds of electric cars, altogether.

  • It's been 24 hours since the last Tesla story!

  • Car moving forward should have three modes:
    1) autopilot - the user need do nothing except enter destination/route and the user can be anywhere in the car, liability for accidents etc is the car manufacture
    2) accident avoidance - the user is in control, but the car will not allow an accident to happen, liability for accidents etc is the car manufacture
    3) manual - user is in complete control and can do damage, liability for accidents etc is the current driver
    This allows for those that want to drive, to be
  • Tesla autopilot is the roomba of self driving cars. It's enough to generate interest in self driving cars and will spur more development till we have something more intelligent.The roomba would blindly go in circles but now we have vacuum cleaners that scan the room and create optimal paths.

    It's a step in the correct direction.

Think of it! With VLSI we can pack 100 ENIACs in 1 sq. cm.!

Working...