TRC is primarily funded by ad revenue. If you like the content you find here, please do not block our ads. Thank you.
Page 23 of 26 FirstFirst ... 132122232425 ... LastLast
Results 331 to 345 of 389

Thread: Driverless car trials

  1. #331
    Should Get Out More
    Join Date
    Jun 2008
    Location
    Milton Keynes
    Posts
    10,195
    Like
    347
    Liked 1,237 in 756 posts
    Blog Entries
    2

    Default Re: Driverless car trials

    Quote Originally Posted by Horse View Post
    It's likely that they will be better than the average non-local driver at knowing where passing spaces are!
    It's also not too much of a stretch to think an AVs will know where all the other cars are in the local vicinity so it'll be able to preemptively wait somewhere, or even just slow down by say 7% so it times things perfectly to arrive at a passing space just when needed.

    Might sound fanciful, but it's not far off what's currently done with air traffic control.


  2. TRC is primarily funded by ad revenue. If you like the content you find here, please do not block our ads. Thank you.
  3. #332
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by Horse View Post
    As I'm sure you also regularly point out, it's often basic things - which have been covered in basic training - which catch people out.
    I don't think there's any doubt that an autonomous car can be programmed to carry out 'routine' tasks, and will undoubtedly be better than the average driver at jobs that require judgement, such as reverse parking - I am still crap at that. They will also help with tasks where there is a risk that the human is driving 'blind' - lane-change tech for example. They can also prevent the kind of driving which 'violates' basic principles such as following distances and excess speed. And they'll be better at controlling a car on a slippery surface.

    However, it's a big leap to extrapolate that to dealing with sudden and unexpected driving emergencies. Let's remember that crashing is actually very rare - many drivers go their entire driving lives without a serious crash. When crashes do happen, they often develop in situations that appear on the surface to be normal but that for some reason take an off-normal path. We talk about 'human error' as if it's something that's easy to detect and correct, but the fact is that as the emergency develops it's usually not obvious that anything unusual is happening. People apply what they know and then find it's not working... and crash.

    As I've said, the roads are essentially uncontrolled compared with a railway... and we still have train crashes. Whilst technology may give the autonomous vehicle better 'senses' and better 'reflexes', it's hard to see how the programmer of an autonomous vehicle can do any better than a human driver at predicting off-normal situations, except by accumulating a huge database based on incidents where the autonomous vehicle DIDN'T perform.

  4. The following user liked this post made by The Spin Doctor:


  5. #333
    Should Get Out More Horse's Avatar
    Join Date
    Dec 2007
    Location
    Always Sunny Central Southern England
    Posts
    16,780
    Like
    1,314
    Liked 1,839 in 1,228 posts

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    it's hard to see how the programmer of an autonomous vehicle can do any better than a human driver at predicting off-normal situations, except by accumulating a huge database based on incidents where the autonomous vehicle DIDN'T perform.
    You may well be right - the difference is that the autonomous vehicle won't be distracted or tired, should be consistent in it's 'observation and planning'.

  6. #334
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by saga_lout View Post
    If you're going to have cars with autopilot deciding to hand over control to the human occupant you'd better make sure that it's fail safe. You'll need a proper handover:
    E.g: "Human must take control in 30 seconds. Acknowledge."
    If no acknowledgement (occupant is asleep or watching a film or something) the car parks itself up and waits for the occupant to wake up or finish watching the film or having a wank.
    You can't just have the autopillot say "That's it, I'm done" and switch off at 70 mph.
    The trouble is that the human who has not been paying attention then has to figure out why the autopilot has switched off... and work out what the necessary course of action is.

    Planes have crashed when that's happened.

  7. #335
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by dodgy View Post
    90% of the journeys my car gets used for are trips between my house and the station. It sits outside my house from 6.30pm until 6am doing nothing, it sites at the station car park from 6.10am to 6.20pm doing nothing. Occasionally I use it at weekends to shuttle kids to sport, but that's it. I really don't need to own it. There are loads of cars with similar usage patterns.

    This could be great for city living too. Just think of how many of the cars parked in inner city streets could be fucked off, leaving that space for something else more human
    I have a car sitting outside here in London. It does just odd journeys, like the one last night to pick some tools up from E Kent, then meet an Aussie instructor just outside Maidstone, then deliver my monthly Biker Down presentation at Rochester. It also goes shopping. So on average it probably moves once a fortnight, maybe less.

    In theory, it's the sort of car that shouldn't be on the streets. And I'd love to get rid of it - in theory.

    In practice, it's worth about 250 so no depreciation to speak of, and costs me around 400 annually to MOT, tax and insure - I put about 20 a month fuel in it on average and other running costs are minimal - it's had a new tyre thanks to pothole damage in the last year, needs new wiper blades (found that out last night) and that's about it - say 60-70 a month.

    How much could I get for my money if I was using a taxi (automated or otherwise) for short trips, or renting a car for the longer journeys?

  8. #336
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by Horse View Post
    You may well be right - the difference is that the autonomous vehicle won't be distracted or tired, should be consistent in it's 'observation and planning'.
    It's also incapable of coming up with a solution on the fly, unlike a human. Human brains are remarkable adept at taking short-cuts in processing data, something that a purely logical machine cannot do - it would have to have been programmed for foresee EXACTLY the situation that's developing in front of it. As I said, airliners have crashed because the autopilot found itself unable to process the incoming data and threw its hands up, tossing the job back into the lap of the unprepared crew.

  9. #337
    Should Get Out More Horse's Avatar
    Join Date
    Dec 2007
    Location
    Always Sunny Central Southern England
    Posts
    16,780
    Like
    1,314
    Liked 1,839 in 1,228 posts

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    It's also incapable of coming up with a solution on the fly, unlike a human. Human brains are remarkable adept at taking short-cuts in processing data, something that a purely logical machine cannot do - it would have to have been programmed for foresee EXACTLY the situation that's developing in front of it. As I said, airliners have crashed because the autopilot found itself unable to process the incoming data and threw its hands up, tossing the job back into the lap of the unprepared crew.
    I think you may well be right. However, there are likely to be work-arounds.



    For example, using the 'single track roads' example mentioned earlier, a geo-fenced L4 car could be driver-driven along the country lanes and L5 vehicles constrained within set zones so they don't go down those lanes at all. Possibly L4s even micro-geofenced (I've made that up, so if you see it elsewhere sometime, I claim the rights ) to areas at certain times of the day, etc.

    It's probably L3 that concerns me the most - and that's the current Tesla-level tech. It works until it doesn't.

    NB https://www.forbes.com/sites/pikeres.../#76b666b84328

    On Aug. 16 2016, Ford held a press conference to announce its plan to launch a fully autonomous vehicle in 2021. Even though the response at the live event was strangely unenthusiastic, there were a number of points that were important for the future of autonomous vehicles and the automotive industry in general.

    The headline news was that in 2021, Ford intends to launch a Level 4 (SAE Standard J3016) fully autonomous vehicle. To clarify the nature of the car, CEO Mark Fields made it clear that it would not have a steering wheel or control pedals, even though last year Ford said it had no plans to sell wheeled pods in which people are merely along for the ride.

    The company also said that it would be several years after 2021 before individuals can buy it; it is aimed at car and ride-sharing fleet operators. Ford Smart Mobility LLC may become one of the first customers. Ford and GM are already piloting their own systems on shuttles for their employees, as noted in a blog earlier this year by my colleague Sam Abuelsamid.

    Ford also said it would continue to develop and improve its driver assistance features up to Level 2 (partial automation), but it would not be introducing any vehicles with Level 3 (conditional automation) because company researchers had concluded that there was no safe way to ensure that drivers would remain alert enough to resume control in an emergency after an extended period of automated driving. Ford vehicles in the future will either have a range of assistance features or be driverless.

  10. #338
    Should Get Out More
    Join Date
    Jun 2008
    Location
    Milton Keynes
    Posts
    10,195
    Like
    347
    Liked 1,237 in 756 posts
    Blog Entries
    2

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    It's also incapable of coming up with a solution on the fly, unlike a human. Human brains are remarkable adept at taking short-cuts in processing data, something that a purely logical machine cannot do - it would have to have been programmed for foresee EXACTLY the situation that's developing in front of it. As I said, airliners have crashed because the autopilot found itself unable to process the incoming data and threw its hands up, tossing the job back into the lap of the unprepared crew.
    This may well be true - however i'd argue that the number of times an autonomous vehicle has 'got it wrong' is still far less than the number of times a human has with the added advantage than a computer only tends to make any given mistake once.

    As I've said, the roads are essentially uncontrolled compared with a railway... and we still have train crashes. Whilst technology may give the autonomous vehicle better 'senses' and better 'reflexes', it's hard to see how the programmer of an autonomous vehicle can do any better than a human driver at predicting off-normal situations, except by accumulating a huge database based on incidents where the autonomous vehicle DIDN'T perform.
    This is also true - but I'd ask: "how did you get into the situation whereby you can predict off normal situations?". The answer of course is that you have experience, which is why new drivers crash more. The downside to your experience is that it's very hard to give it to someone else. You can teach, but you can't directly pass on everything you understand about driving to another person - which is exactly the problem you're describing for a computer. The advantage a computer has over you is that it can pass it's experience directly over to someone else. I totally agree it'll take miles to come up with truly 'inventive' AVs, which is why they're not deployed en mass straight away.

  11. #339
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by Horse View Post
    Ford vehicles in the future will either have a range of assistance features or be driverless.[/B]
    I can see a role for driverless vehicles doing shuttle loops along predetermined paths... although trams do the same.

  12. #340
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Your friendly local advanced trainer :)
    Posts
    7,865
    Like
    476
    Liked 567 in 344 posts

    Default Re: Driverless car trials

    Quote Originally Posted by Mr. Dazzle View Post
    This may well be true - however i'd argue that the number of times an autonomous vehicle has 'got it wrong' is still far less than the number of times a human has with the added advantage than a computer only tends to make any given mistake once.
    That does suppose that the programmer identifies the problem correctly, then comes up with the solution and that the vehicle is capable of carrying out the programming.

    Given that millions of cars are in use in the UK alone, what we'll see is not a mature system hitting the roads but a series of developmental stages being rolled out one after the other. Whilst the manufacturer will certainly retain the ability to roll out updates, I would expect them to be limited when it comes to any particular model - that way new features will be retained for new vehicles.


    This is also true - but I'd ask: "how did you get into the situation whereby you can predict off normal situations?". The answer of course is that you have experience, which is why new drivers crash more. The downside to your experience is that it's very hard to give it to someone else. You can teach, but you can't directly pass on everything you understand about driving to another person - which is exactly the problem you're describing for a computer. The advantage a computer has over you is that it can pass it's experience directly over to someone else. I totally agree it'll take miles to come up with truly 'inventive' AVs, which is why they're not deployed en mass straight away.
    Absolutely... but the human brain is far more flexible. The roads are insanely complex... just watch any speeded-up video of a busy roundabout. We're not just monitoring where other vehicles are but we're making informed guesses about what other road users are going to do, plus dynamic risk assessments based on our chances of getting away with what we want to do.

    Given that few people really understand why they crashed, it seem to me it'll be a job and a half for a programmer several times removed from a real-life crash to feed in the right instructions to avoid a repeat.

    When a 21st century autopilot which only has to keep an incredibly expensive plane flying straight and level based on a very limited number of parameters fails, I really do struggle to see how the kind of low-cost piece of kit that will find its way into a mass-produced car will cope, particularly when we have a mix of human control, cars with driver assist and genuine autonomous driving, all in close proximity. I'm by no means an aviation expert but I believe some of the most dangerous situations have developed when human-flown planes have come into close proximity with airliners.

    One of the things to consider is that a mooted advantage of autonomous vehicles is their theoretical ability to run in cohorts - the idea is that the lead vehicle controls all the others and so you can close them right up and fit more on the road in the same space. The downside is obvious - if the lead vehicle gets it wrong, you've just guaranteed a pile-up. We may well cut out the really daft crashes... but I suspect that when they do occur, they will be far more serious... rather like what happened to the railways in the last century. Train crashes are pretty rare... but tend to be bad ones when they do happen.

    I suspect it'll be absolutely necessary for vehicles to talk to each other as well as their environment and for their movements to be under full computer control using some kind of net which connects everything on the roads. I know there's research going in that direction. I may turn out to be wrong, but I doubt we can ever have an independently autonomous vehicle that is 100% capable.

    Horse said that the enthusiasm level for fully autonomous cars was underwhelming. As I enjoy driving, I can't say I'm surprised.

  13. #341
    Not Much To Do
    Join Date
    Dec 2014
    Posts
    906
    Like
    154
    Liked 308 in 192 posts

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    Horse said that the enthusiasm level for fully autonomous cars was underwhelming. As I enjoy driving, I can't say I'm surprised.
    The perspective is so important, and is likely to influence judgements in most people.

    I, on the other hand, look around at drivers and think, 'this is mad.' My mother has several friends in their nineties who drive, without recent training, and she thinks some are a menace owing to very slow reactions and low awareness. They do crash sometimes. At other times one looks through a car window to see that the vehicle is controlled by a teenage boy, sometimes one who appears to project alarming aggresion.

    The sanity threshold for legal control of a motor car is considerably lower than that required for legal ownership a gun designed to take pot shots at rabbits, so naturally motor vehicles are becoming a weapon of choice among terrorists, as well as use by ordinary people who once in a while feel the need to put cyclists in their place, and so on. The level of skill required to drive a car legally is low--perversely it is lower than is required for a motorcycle. For these reasons, I'd be happy to see fewer humans in control of cars. The fact that many drivers enjoy driving, including no doubt the Toad-of-toad-hall nongenarians, overheated schoolboys, mass murderers on a short-cut to heaven, and otherwise sane drivers who use risky overtaking to bolster their sense of self-worth, should not be allowed to skew policy, though for political reasons it might.
    Last edited by John B; 13-09-17 at 14:53.

  14. #342
    Should Get Out More Horse's Avatar
    Join Date
    Dec 2007
    Location
    Always Sunny Central Southern England
    Posts
    16,780
    Like
    1,314
    Liked 1,839 in 1,228 posts

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    I can see a role for driverless vehicles doing shuttle loops along predetermined paths... although trams do the same.
    Already happening. Trials in several locations, yes, rather than a regular service, but happening all the same.

    Quote Originally Posted by The Spin Doctor View Post
    Given that millions of cars are in use in the UK alone, what we'll see is not a mature system hitting the roads but a series of developmental stages being rolled out one after the other.

    When a 21st century autopilot which only has to keep an incredibly expensive plane flying straight and level based on a very limited number of parameters fails , I really do struggle to see how the kind of low-cost piece of kit that will find its way into a mass-produced car
    Already happening. The Google cars, Teslas, Ubers and many more are on the roads, including the UK's. Economy of scale etc. will be the key to getting the tech in vehicles. We bought an Ibiza in about 2001 that was s.h then but had drive-by-wire throttle for the FI and electronic ignition. There are 'cheap' consumer cars running around with all sorts of tech built-in.

    Don't forget the figures I posted very recently of the vast sums being spent in this field.

    Quote Originally Posted by The Spin Doctor View Post
    I believe some of the most dangerous situations have developed when human-flown planes have come into close proximity with airliners.
    I think I said earlier that this sort of 'mixed' driving is being researched too.

    Quote Originally Posted by The Spin Doctor View Post
    Train crashes are pretty rare... but tend to be bad ones when they do happen.
    Same with plane crashes? But are they actually relatively 'dangerous' modes of transport?

    Quote Originally Posted by The Spin Doctor View Post
    I may turn out to be wrong, but I doubt we can ever have an independently autonomous vehicle that is 100% capable. .
    I doubt whether we have any drivers who are 100% capable . . .

    Quote Originally Posted by The Spin Doctor View Post
    Horse said that the enthusiasm level for fully autonomous cars was underwhelming. As I enjoy driving, I can't say I'm surprised.
    Probably 'quoted' rather than 'said'

    Check out the yoof, though - they're becoming less interested in owning mobility. It could be that the era of 'bloke in a shed' etc is going away . . .

    Quote Originally Posted by John B View Post
    My mother has several friends in their nineties who drive
    At F-i-Laws flat complex, there is a lovely, active, lady in her 90s. You can see a photo of here graunched Corsa 'Elegance' in the photos thread. That's now fixed, but she's gained another on the front . . .

  15. The following user liked this post made by Horse:


  16. #343
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Down in the Duchy (at last)
    Posts
    14,015
    Like
    3,053
    Liked 1,725 in 1,211 posts

    Default Re: Driverless car trials

    Quote Originally Posted by Mr. Dazzle View Post
    It's also not too much of a stretch to think an AVs will know where all the other cars are in the local vicinity so it'll be able to preemptively wait somewhere, or even just slow down by say 7% so it times things perfectly to arrive at a passing space just when needed.

    Might sound fanciful, but it's not far off what's currently done with air traffic control.
    And it could work too.

    Of course it will need a major re-mapping of the UK, since even OS maps don't show which bits of narrow road are wide enough to pass, and which bits aren't.

    Unless AVs always reverse to somewhere they have passed and measured as wide enough. That would be acceptable (to us that don't own AVs)

  17. #344
    Should Get Out More
    Join Date
    Jan 2008
    Location
    Down in the Duchy (at last)
    Posts
    14,015
    Like
    3,053
    Liked 1,725 in 1,211 posts

    Default Re: Driverless car trials

    Quote Originally Posted by The Spin Doctor View Post
    The trouble is that the human who has not been paying attention then has to figure out why the autopilot has switched off... and work out what the necessary course of action is.

    Planes have crashed when that's happened.
    And pilots typically have to re-qualify in a simulator every 6 months or so, when weird situations that the automatic systems can't handle are thrown at them to see how, and how quickly, they react. Can you imagine DVLA re-qualifying drivers every 6 months .............
    Planes also typically have slower responses, at 30,000 feet you do often have 10s of seconds to respond to an emergency. In a car it is more like 1/10ths of a second.

  18. The following user liked this post made by Cousin Jack:


  19. #345
    Should Get Out More
    Join Date
    Jun 2008
    Location
    Milton Keynes
    Posts
    10,195
    Like
    347
    Liked 1,237 in 756 posts
    Blog Entries
    2

    Default Re: Driverless car trials

    Quote Originally Posted by Cousin Jack View Post
    And it could work too.

    Of course it will need a major re-mapping of the UK, since even OS maps don't show which bits of narrow road are wide enough to pass, and which bits aren't.

    Unless AVs always reverse to somewhere they have passed and measured as wide enough. That would be acceptable (to us that don't own AVs)
    Even today Google posses aerial photographs of every inch of the UK - save a few places like military bases - down to a resolution of a metre or two. How much better do you think that will be in 20 years time, given that in 2017 there are private companies that can photograph every inch of the planet literally every day. Also how much better do you think it will be when every car is constantly mapping and re-mapping the road?

    I think people are underestimating how fast computers have developed and how much data is out there today, let alone in years to come.

  20. The following user liked this post made by Mr. Dazzle:


Page 23 of 26 FirstFirst ... 132122232425 ... LastLast
Go Back to Forum My Forum