How will we drive cars in the future?

Incase that comment was not ment being sarcastic, I think its bull. There are life supporting medical machines, and not just a few, that run on windows based software too. I doubt they would use software that easily crashes on such machines. Besides that its not like other OS are that perfect. I used linux and novell and I always managed to crash them way faster than any windows OS. Especially novell, it seems novell has the tendency to just totally melt down when you reboot the server...
 
Incase that comment was not ment being sarcastic, I think its bull. There are life supporting medical machines, and not just a few, that run on windows based software too. I doubt they would use software that easily crashes on such machines. Besides that its not like other OS are that perfect. I used linux and novell and I always managed to crash them way faster than any windows OS. Especially novell, it seems novell has the tendency to just totally melt down when you reboot the server...

The fact that you see GUIs of medical equipment running on Windows or Linux is irrelevant. No life supporting or otherwise safety critical system is allowed to rely on Windows/Linux working as assumed. The safety functions must be performed by other means ( another processing unit running software that can actually be shown to be safe, or electrical or mechanical fallbacks).

Neither Windows nor Linux variants are accepted (and surely not intended) for such tasks.
 
What will happen is that you'll see auto-drive with manual override, but in the event of a crash the responsibility by law will still be on the driver. There will be big warning about how you still can't take your eyes off the road. Eventually the systems will get better and better, possibly with car-city communication to help aid traffic flow, and maybe one day the statistics wil make insurance agencies actually recommend (via discounts) the use of complete automation. They'll even be able to completely recreate the scene around a traffic accident.

It will happen. :devilish:
 
When you're talking about medical apps, the main thing that makes that possible is the apps themselves and not the OS they're running on. It's not so much that Windows is a self-destructive OS -- it's everything that people do with (and to) Windows. Any OS can get pushed into a corner where something wrong happens or the OS simply doesn't handle a condition very robustly. Because Windows is so much more widely used, there are just that many more ways to fall into that sort of hell. And the reason it's viewed as so unstable is simply the fact that we run into problems like that doing otherwise mundane things -- meaning that either the software itself is crap, or the OS' ability to handle that software and all the things it does is crap, or a combination of both.

In the case of control software for engine timing/ignition, traction control, ABS, etc., that's comparatively simple, and hence pretty robust stuff. There's no "OS" in the way we'd think of it, and most everything is more or less standalone -- the really complicated bit is the vast array of sensors which are needed to provide reliable information to the software. It's very unlikely that such software would crash unless there was something screwy with the information received and yet the sensor sending this data doesn't register as a failure -- which itself is a pretty unlikely thing considering most of these sensors either work just fine or just plain fail completely. The latter condition is pretty safe other than the fact that a system might effectively be disabled or work in a sort of "default" mode depending on what it is.

Self-driving systems, otoh, are a whole other ball of wax, because you're talking about relying on unreliable information (or at least insufficient information to make a reliable analysis). It's very easy to fool these sorts of things, and the machine's "awareness" of its place makes the whole thing really prone to screwing up. Even those self-parking vehicles typically depend on an analysis or verification of the space from the driver in order to work properly. If it screws up in the analysis of the image, it screws up the parking job. The same thing can happen when driving at highway speeds, and then what? You've basically killed a family. It's still loaded with unsolved problems even in the academic space. There's a method for handling this, a method for handling that, and a million others and if you think that combining them all will make it all robust, you're in for a huge surprise, and surprises in engineering are always bad.

In an age where you have in-wheel motors and such, it is a given there will be at least a certain extent of change given the absence of tranmissions and driveshafts and so on. Analog pedal controls and an otherwise ordinary steering wheel and so on are straightforward from an input standpoint. Proper feedback mechanisms that give the driver information of the same quality they'd get from a more genuine link to the mechanisms is a little harder. Not so much that it can't be done, but that there will always be that "fakeness" to it. In a lot of respects, it might be easier to just build these sorts of new vehicles with about the same level of mechanical/fluid linkages they currently have. From an engineering standpoint, this is also a safer bet, I would think. I know we're already at the point where the gas pedal is completely dbw in a lot of cars -- it's a remote possibility that people will just get used to it, but it certainly won't happen soon. I'd rather feel my steering wheel bounce when I go over a bump on one side than just assume everything will behave when I shift the handle positions (a la GM Hy-Wire controls).
 
Last edited by a moderator:
Some kind of automated driving may come eventually, but it will be a long time until then and manual controls will still be required. I'm really more interested in the mid-term prospects. The Hy-Wire concept may not be perfect controls-wise, but at least the design team dared to challenge conventions, and I'm pretty certain some positive change will come out of it.
 
Self-driving systems, otoh, are a whole other ball of wax, because you're talking about relying on unreliable information (or at least insufficient information to make a reliable analysis). It's very easy to fool these sorts of things, and the machine's "awareness" of its place makes the whole thing really prone to screwing up. Even those self-parking vehicles typically depend on an analysis or verification of the space from the driver in order to work properly. If it screws up in the analysis of the image, it screws up the parking job. The same thing can happen when driving at highway speeds, and then what? You've basically killed a family. It's still loaded with unsolved problems even in the academic space. There's a method for handling this, a method for handling that, and a million others and if you think that combining them all will make it all robust, you're in for a huge surprise, and surprises in engineering are always bad.

However valid that analysis may be, we already *know* that humans aren't robust. I think that the GPS is the first step in a long line of innovations that will end with cars that can reliably (and more reliably than most humans) drive themselves. And personally, I wish they were already here. You could completely redesign the interior of a car and make it a comfortable working area for instance. Often the car is still a good or the best way to travel for me (although I take the train a lot more than any of my colleagues already), but I always find the four or more hours I have to spend in a car a complete and utter waste of time. In the train, I usually manage to get even more work done than in the office (less distractions)

Modern cars are dangerous by the way - I was in a new Nissan Quashkai II (or however you spell that) and you have no sense of speed at all. It's a very good car, but it's like a lot of other modern cars (Audi) - they invite you to drive extremely fast and you barely notice that you do. Long time ago already my father said he liked driving a mini cooper, because even at 100km/h you get a sense you are going really, really fast. :D
 
And what about all the camera's and sensors? also pure production cost isnt everything that matters. R&D will cost alot of money and that money needs to be recouped one way or another.
 
Oh, there's nothing wrong with them per se. But how do we know there's nothing better without even trying? Sometimes it's necessary to challenge preconceptions, or you end up with a suboptimal solution just because people are familiar with it. Take the QWERTY keyboard as an example.

Maybe pedals are indeed ideal, but I don't think electric cars should try to mimic the pedal response of ICE cars.



There's little evidence that qwerty is particularly suboptimal compared to any other keyboard layout scince no one's ever done any proper studies into the area.

And how do you mean not mimic the pedal response of ICE cars? Pretty sure accelerators aren't mechanically attached in fuel injection cars anyway, it just provides an input for the ECU. I'm not sure what you could do to the brake either. You press it more you stop faster, you can't remove the analogue nature of either, might work for arcady race games but in reality you're just going to wheelspin and emergancy stop everywhere, not to mention driving on snow/ice would be interesting, or the outcome on passenger comfort.

You could move both onto the steering wheel but, I'm not convinced on the twist/grip accel/brake system, could get a little tiring on the hands after a while and maintaining the same accel rate while you go round a corner would demand finer control and better coordination than with pedals
 
There's little evidence that qwerty is particularly suboptimal compared to any other keyboard layout scince no one's ever done any proper studies into the area.
Depends on what you call "proper studies". But given that there's also little research speaking in favour of qwerty it's extremely unlikely that it ranks among the best possible layouts.

And how do you mean not mimic the pedal response of ICE cars? Pretty sure accelerators aren't mechanically attached in fuel injection cars anyway, it just provides an input for the ECU. I'm not sure what you could do to the brake either. You press it more you stop faster, you can't remove the analogue nature of either, might work for arcady race games but in reality you're just going to wheelspin and emergancy stop everywhere, not to mention driving on snow/ice would be interesting, or the outcome on passenger comfort.
I don't know how common ETC is exactly, but there still are a lot of cars around without. But even then the "accelerator" doesn't control acceleration, it controls the throttle of an engine with a highly non-linear torque output.
I'm not at all suggesting getting rid of the analogue nature. Just that creating familiarity for drivers of ICE cars shouldn't be the main goal when fine-tuning the response curve of drive-by-wire controls.
 
However valid that analysis may be, we already *know* that humans aren't robust. I think that the GPS is the first step in a long line of innovations that will end with cars that can reliably (and more reliably than most humans) drive themselves.
My point was more that the level they're at right now is still light years away from even being close to the robustness of human beings -- and the problems that plague them aren't just unsolved at the practical level either; They're unsolved even at the level of strictly academic exercises.

The cases where they handle themselves really well are pretty trivial, and a person who isn't asleep at the wheel or heavily inebriated would not even have to put in any effort to do better. There's still the problem of acquiring sufficient information. There are things like the VW autonomous Golf, and that works for the case of road courses and such after memorizing a path, but it certainly isn't built to dynamically adapt to things like traffic and detour routes and so on. Most all such systems will require augmentation from the environment as well such as traffic lights actually feeding signals back to the cars themselves.

If you're using tools like this to augment the limits of what a human being can do, that's fine. And there are several cases where this sort of thing already exists. Cars that have smart cruise control that senses traffic, self-parking, night-vision windshield HUDs, crash alerts and so on. All this stuff can *help*, but don't kid yourself that we'll be able to have autonomous cars within the next 20 years. It's simply nowhere near that point.

Bear in mind that with anything like that, the kinds of safety standards and rigorous testing and approvals and several tonnes (literally) of red tape you have to go through means that even if, by some miracle, we solved it all within 2009, it'll still be 2020 before it's even allowed to hit a single production vehicle.

Modern cars are dangerous by the way - I was in a new Nissan Quashkai II (or however you spell that) and you have no sense of speed at all. It's a very good car, but it's like a lot of other modern cars (Audi) - they invite you to drive extremely fast and you barely notice that you do.
That I can't really picture being an issue in the cases where a self-driving computer wouldn't work out. When I was in college, the 225-mile drive back home for any holiday period often caught me driving well over 90 without realizing it. It's not so much that the car was failing to exude any sense of speed, but that the highway was 4 lanes and almost completely empty all the time. The sort of vastness of empty space all around you would basically mask just how fast you were going. The same thing never comes close to happening when you're in a tighter

computers are cheap. hell GPS are under 100 euros now!
As long as you buy one without a screen, anyway. Difference in price between a simple receiver and one with software that actually gives you a map display and all is anywhere from 250 to 700 USD.

And how do you mean not mimic the pedal response of ICE cars? Pretty sure accelerators aren't mechanically attached in fuel injection cars anyway, it just provides an input for the ECU.
Depending on the vehicle, it could be mechanically linked to the throttle body valve. Many common vehicles these days do use a drive-by-wire gas pedal as you say, but when you get into higher-end or more enthusiast-oriented vehicles, I think the direct throttle body linkage is still there. There are exceptions, of course, particularly with all the borrowed tech that goes around, but that just goes with the territory.
 
Depends on what you call "proper studies". But given that there's also little research speaking in favour of qwerty it's extremely unlikely that it ranks among the best possible layouts.


proper being independent, pretty much any study on the subject is written by someone with a bias either to proving that their solution is better than qwerty, or that qwerty is no worse than someones solution. I'm sure qwerty probably isn't the optimal for most, it was designed to separate the majority of successive keystrokes (in english) to not be on the same vertical line as far as typewriter hammers were concerned, but it's also unlikely to be measurably suboptimal. It probably doesn't help that the optimal layout will vary from person to person based on a whole variety of factors such as the relative strength/dexterity of their fingers and even their own vocabulary. Given though that the vast majority of computer users unfortunately never progress beyond using about 3 fingers and anyone who does can change their keyboard layout to anything they like with a screwdriver and some free software, i guess most of this is largely irrelevant.

But that's a similar analogy to cars, to the vast majority of people it's a car, providing it gets them from A to B, they can get the shopping in the back, fit the kids in etc they don't tend to be that bothered. The response and feel of accelerators already varies quite a bit between different cars and even different engines within the same car. Steering and gearboxes feel also varies greatly between vehicles yet many people won't really notice the difference and hardly any will actually be interested or bothered about it. The Focus is the most bought car year on year in the UK, yet I'd bet if someone did a survey of focus owners most either won't know or won't have bothered investigating that you can alter the power steering feel.

To me that presents a problem for R&D in manufacturers, when you're trying to reinvent the control systems for the average person (ie most of your customers) and the average person really isn't bothered.

I guess one area that will be investigated, looking at the limited range of motion on the hywires steering, probably already has been, is making the wheels turn more at slower speeds allowing for more accurate fine steering at high speeds and easier manoeuvring at lower speeds. However it might be better to put that as a manual thing, having the steering rate change as you accelerate through a corner while maintaining the same steering 'wheel' angle might prove undesirable, and the racy types might complain at the lack of deliberate over/reverse steering at speed.


That I can't really picture being an issue in the cases where a self-driving computer wouldn't work out. When I was in college, the 225-mile drive back home for any holiday period often caught me driving well over 90 without realizing it. It's not so much that the car was failing to exude any sense of speed, but that the highway was 4 lanes and almost completely empty all the time. The sort of vastness of empty space all around you would basically mask just how fast you were going. The same thing never comes close to happening when you're in a tighter
The car does make a large difference too, my own car is a small city one with slightly shy of 70 horses and it's pretty obvious when you're doing 90 in that. Driving a mid sized saloon with reasonable power (eg 150hp) and 6 gears and I've found myself doing as much as 20mph more than I was intending even on busier motorways. I'm not sure it's particularly more dangerous though, I've never found the same issues when corners are involved, bigger cars, while not giving such a sense of speed do give a sense of weight quite well making it quite clear you aren't going to make it around that corner at a certain speed.
 
I would feel much safer if all cars were reliably driven by computer, instead of temperamental humans that might drift off and start doing anything other than driving a car. The key word being "reliably".

But, I agree with ShootMyMonkey: most "error correction" in programs is done by raising an error message. Which is real fun, if that means that your car, doing 90, just became a dumb, unguided missile.

And if the error isn't raised, that generally means the program doesn't noticed that something just went seriously wrong. With the same result.

Cars have no spatial awareness whatsoever. And they really need that, to be able to drive independently. It's one of those "hard" problems.
 
there's a slight problem with the premise given in the title, I should have seen it :).
hopefully we won't drive cars in the future. they should be reserved to people with ill health in rural areas, law enforcement, medic etc.

imo you will pedal (the passenger may pedal depending on the type of vehicle) with an electrical assistance, on a variety of 2-, 3- and 4-wheeled light vehicles. enough for common trips, picking up groceries etc. Longer range travel by rail, bus, boat etc.
 
Last edited by a moderator:
Back
Top