Automated Driving

  • I don't know if we'll ever get there, but it wouldn't surprise me. However, I don't think we'll get there by taking a big leap the way most people seem to think about it. If we get there it will be by dribs and drabs, they way most large systems that actually function get 'designed'.

    We already have vehicles that can warn or even take action when confronted with situations like tailgating. We have vehicles with systems that help maintain control under slippery conditions and some are 'skill-competitive' with serious amateur performance drivers like people that ice-race for a hobby. I can easily imagine the day when those systems are skill-competitive to people that ice-race for a living. Others here have mentioned self-parking cars and cars that hold a lane at least temporarily if the driver nods off. Okay, Volvo blew the 'pedestrian saver', but somebody might yet figure it out.

    I guess what I'm trying to say is that it looks to me like we're already building the system one piece a time.

  • A big advantage of cars that drive themselves would be that people who are currently unable to drive would be able to use a car: physically or mentally handicapped, blind, young children, etc.

    A self-parking car would eliminate a lot of problems, like finding a parking spot within a reasonable walking distance from your destination, and having spots reserved for the handicapped.

    "Car, drop me off at the theater, return home, and then pick me up in 3 hours."

  • .

  • GSquared (11/10/2010)


    Kevin Hood (11/10/2010)


    The true test of an automated driving system will be how it responds when a squirrel runs onto the road.

    Or someone's kid.

    Or knowing the difference. A system that responds the same way when a squirrel runs across the road as when a child does is a problem.

    ...

    -- FORTRAN manual for Xerox Computers --

  • Some interesting "get there from here" possibilities are coming to mind. Imagine that we allow self-driving cars, but only for < 15mph. Perhaps you could have cars that dropped you off at central places like stores/malls/work, and then they car "self parked". It would certainly be more efficient than building some of those large automated parking garages that have to move the cars themselves.

    I can see that a car responding to a squirrel and a kid the same in a parking lot at 10mph isn't so bad. At 25-50mph, it's a much bigger deal.

  • jay holovacs (11/12/2010)


    GSquared (11/10/2010)


    Kevin Hood (11/10/2010)


    The true test of an automated driving system will be how it responds when a squirrel runs onto the road.

    Or someone's kid.

    Or knowing the difference. A system that responds the same way when a squirrel runs across the road as when a child does is a problem.

    Less traumatically, a system that responds to a pedestrian jaywalking the same way as to a cop directing traffic in an intersection is going to be problematic.

    Lots of issues that require both visual identification and judgement are going to be very, very hard to code for.

    - Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • Steve Jones - SSC Editor (11/12/2010)


    Some interesting "get there from here" possibilities are coming to mind. Imagine that we allow self-driving cars, but only for < 15mph. Perhaps you could have cars that dropped you off at central places like stores/malls/work, and then they car "self parked". It would certainly be more efficient than building some of those large automated parking garages that have to move the cars themselves.

    I can see that a car responding to a squirrel and a kid the same in a parking lot at 10mph isn't so bad. At 25-50mph, it's a much bigger deal.

    Okay, now we're getting somewhere! Pull that top speed down to 5 mph and the object detection gets a lot simpler and it probably has minimal impact on the time to park/retrieve the car. And pedestrians (and squirrels!) can probably take on the primary responsibility for not getting hit (bumped?), especially if there also paths that are off-limits to cars.

  • Ron Porter (11/12/2010)


    Steve Jones - SSC Editor (11/12/2010)


    Some interesting "get there from here" possibilities are coming to mind. Imagine that we allow self-driving cars, but only for < 15mph. Perhaps you could have cars that dropped you off at central places like stores/malls/work, and then they car "self parked". It would certainly be more efficient than building some of those large automated parking garages that have to move the cars themselves.

    I can see that a car responding to a squirrel and a kid the same in a parking lot at 10mph isn't so bad. At 25-50mph, it's a much bigger deal.

    Okay, now we're getting somewhere! Pull that top speed down to 5 mph and the object detection gets a lot simpler and it probably has minimal impact on the time to park/retrieve the car. And pedestrians (and squirrels!) can probably take on the primary responsibility for not getting hit (bumped?), especially if there also paths that are off-limits to cars.

    They have been doing that for years now with the new assisted breaking feature in cars. Any object detected within a safety zone at high speed, alerts the breaks to auto apply.

  • I still think it's all "pie in the sky". Whilst there is technology that can handle the individual bits, like parking & keeping a given distance from another car etc I still believe that there are just too many outside influences that could not be programmed for and be compatible.

    We have weather conditions that can vary from minute to minute and also height & location. I travel 30 miles to work and experience 3 types of weather.

    We have squirrels & people's kids bears/deer/buffalo/elk etc depending country. You might not want to avoid a squirrel but you'd do a lot of damage if you hit one of the big furry things mentioned.

    We have drunks wondering onto the road.

    We have things falling on the road (trees, cement mixers etc)

    We have mechanical failures plus the fall out I had a radiator bust when a bit fell off the scrap wagon in front of me.

    We have all the different kinds of traffic (push bike, motor bikes, classics, tractors, vans, lorries, combined harvesters, crop sprayers).

    Then there are the people like me who actually ENJOY driving.

    Any system like this would have to be a closed circuit

  • Mick Moses (11/12/2010)


    I still think it's all "pie in the sky". Whilst there is technology that can handle the individual bits, like parking & keeping a given distance from another car etc I still believe that there are just too many outside influences that could not be programmed for and be compatible.

    We have weather conditions that can vary from minute to minute and also height & location. I travel 30 miles to work and experience 3 types of weather.

    We have squirrels & people's kids bears/deer/buffalo/elk etc depending country. You might not want to avoid a squirrel but you'd do a lot of damage if you hit one of the big furry things mentioned.

    We have drunks wondering onto the road.

    We have things falling on the road (trees, cement mixers etc)

    We have mechanical failures plus the fall out I had a radiator bust when a bit fell off the scrap wagon in front of me.

    We have all the different kinds of traffic (push bike, motor bikes, classics, tractors, vans, lorries, combined harvesters, crop sprayers).

    Then there are the people like me who actually ENJOY driving.

    Any system like this would have to be a closed circuit

    The thing on that argument is, though, that human drivers often don't deal with those situations very well either. The software doesn't have to be perfect, it just has to be slightly better, overall, than a human driver. If it's just as bad at dealing with deer/kids in the road as human drivers, but doesn't get drunk, then it's a potential improvement. If it also costs a million dollars per car, then it's not an improvement at all. And so on, comparing features/costs at each point.

    As for the enjoying part, which will generate more fun: Driving, or hacking your Auto-Car (TM) so that it drives you to work backwards just for the heck of it? Or hacking it so it does serious donuts in the WalMart parking lot? Or hacking your rental car/shared car, so it drives through red light districts no matter what destination and starting point the next driver asks for? The possibilities are endless!

    If you think loud stereo systems are annoying, you haven't imagined the possibilities for hackers with automatic cars!

    - Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • It’s just a matter of time until it is possible. Faster processors, better sensors, more sophisticated algorithms are all being worked on now, and the financial incentive of a huge world wide market ensures progress will continue.

    It doesn’t have to all happened at once; just a steady introduction of improvements over time that make it incrementally better. We are already getting there with traction control, computer controlled engine and transmission, collision detection, and automatic parking.

    How about when cars get to be much better than a human can do now? Imagine safe driving at 200 mph at night on rain-slick highway with satellite navigation and road condition updates for automatic selection of optimal routes while the passenger catches a nap, watches TV, or posts on SQL Server Central. Is there anyone that wouldn’t want that? “We’re leaving Denver right now; we should be in Kansas City in about 3 hours.”

  • Michael Valentine Jones (11/12/2010)


    ... “We’re leaving Denver right now; we should be in Kansas City in about 3 hours.”

    Where are the flying cars I was promised?

  • Adam Gardner (11/12/2010)


    Michael Valentine Jones (11/12/2010)


    ... “We’re leaving Denver right now; we should be in Kansas City in about 3 hours.”

    Where are the flying cars I was promised?

    The people who were going to invent them were killed in car crashes.

    - Gus "GSquared", RSVP, OODA, MAP, NMVP, FAQ, SAT, SQL, DNA, RNA, UOI, IOU, AM, PM, AD, BC, BCE, USA, UN, CF, ROFL, LOL, ETC
    Property of The Thread

    "Nobody knows the age of the human race, but everyone agrees it's old enough to know better." - Anon

  • Michael Valentine Jones (11/12/2010)


    It’s just a matter of time until it is possible. Faster processors, better sensors, more sophisticated algorithms are all being worked on now, and the financial incentive of a huge world wide market ensures progress will continue.

    It doesn’t have to all happened at once; just a steady introduction of improvements over time that make it incrementally better. We are already getting there with traction control, computer controlled engine and transmission, collision detection, and automatic parking.

    How about when cars get to be much better than a human can do now? Imagine safe driving at 200 mph at night on rain-slick highway with satellite navigation and road condition updates for automatic selection of optimal routes while the passenger catches a nap, watches TV, or posts on SQL Server Central. Is there anyone that wouldn’t want that? “We’re leaving Denver right now; we should be in Kansas City in about 3 hours.”

    It's much to easy to over estimate how well machines do things. In fact, they can handle very specific, very limited circumstances fairly well. But the real world is far messier. A decent driver is looking at a lot of stuff... whether the driver approaching the stop sign appears to be paying attention (or is on the phone), the deer standing in the brush, the kids playing in a field. The nuances in the world around us, nuances that we as humans don't even conciously think about but are completely unavailabe to our sonar equipped machines, make most of the difference. Unfortunately the sonar fingerprint for a dangerous situation and a minor one is incredibly ambiguous. While sometimes a pedestrian near a moving car is a danger signal, there are also plenty of times when a car DOES legitimately navigate though a field of pedestians where eye contact between the driver and the pedestrian is the key factor.

    For all the hoopla, there is still not a robot made that can ride a bicycle or motorcycle. Watch a squirrel jump from branch to branch, judging the distance, orienting himself in flight... there is no machine that can come close to that. Machines do limited things very well, they can play a good game of chess, but can't conceive a business plan.

    A while back I read an article (if I find the link, I'll post it) by a truck driver about some of the computerized 'safety' functions being required on new trucks. Some were helpfull (ABS to prevent locking the trailer wheels), but some of them were truly not ready for prime time. Lane change warnings that constantly screamed warnings and went nuts when on a road with no fog line (and couldn't be switched off). Anti jacknife systems that got confused by potholes and misapplied the brakes.

    Automatic systems do not see what's going on, they only have a sonar fingerprint. They don't hear what's going on, they understand the difference between a dangerous situation and a harmless one.

    ...

    -- FORTRAN manual for Xerox Computers --

  • I think you're missing one point, but when all the cars on the road are robots, they'll be far more predictable, and that is one large load of hazards that won't need to be catered for. I do agree that this won't cater for animals and people crossing the road etc, but I'm sure there are ways of reorganising the roads to cater for this.

Viewing 15 posts - 46 through 60 (of 75 total)

You must be logged in to reply to this topic. Login to reply