Why driverless car's are a LONG way off.

[witko999]

technodup said:

Pool cars would mean fewer cars and more efficient usage so I don't see that. Plus rush hour congestion is partly caused by humans ignoring speed limits/signals, poor lane discipline, rubbernecking, last minute changes of direction etc. Take that away and the situation would be very different.

Plus I can see different levels of car, similar to taxis, private hire, limo etc we have today. Want the no frills, hosed down at the end of the day 'public transport' option? £50/m. Want the exclusive leather seats and full entertainment system, reserved for those with taste? £200/m please.

The idea that the market, the manufacturers and the new companies that will emerge won't have solutions to these 'issues' is naive.

We have the option of car pooling right now but it's seldom used. In the future, why would anyone want to wait for a car to show up, then sit with 3 strangers while the car takes them to 3 other peoples place of work before finally arriving at their own? You may as well take the bus.
People enjoy having their own vehicle and not having to be inconvenienced. Also, think of how messy and disgusting your average bus is at the end of the day, and then imagine what a driverless pool type car would look like after running people around all day. Discarded food wrappers, drinks etc. I'd want plastic overalls just to sit in it.

Of course, none of these problems are insurmountable, but to think it's coming anytime soon is what's naïve.

Regards the congestion, driverless cars are not going to magically make huge road capacity increases.

[deckster]

witko999 said:

Of course, none of these problems are insurmountable, but to think it's coming anytime soon is what's naïve.

Of course this is nub of the matter - what do we mean by 'soon'? I absolutely agree that it's not going to happen this year, or this decade. By 2030? Possibly. In my lifetime (being in my early 40s)? Very likely. In my kid's lifetimes? Without a doubt.

I will make a firm prediction now that that my children's generation will be the last to own driving licenses.

[otolith]

EnglishTony said:

I am going to have to disagree with you, sorry, I'm not doing just to pass the time or because I like disagreeing with people.

There are other issues at play here. If these things are truly driverless, assuming you don't need to be sober to tell if where to go and notice when it's got it wrong and is taking you to your ex girlfriends rather than your current one's place, the more serious point is about insurance.

"Hello, my name is English Tony and I'd like to insure a driverless car please and yes I intend to get in it rat-arsed". That'll be a cost problem then.

Alternatively Elon Musk and his ilk as manufacturers will have to carry the Insurance burden. For a vehicle that takes unsupervised drunks from a to b. These costs will obviously be passed on to the customer.

I really don't see why this is complicated by autonomy - you insure the car against third party risks resulting from the use of it on the road. It makes no difference whether someone is manually driving it or not. It's not about who is at fault, it is about who is liable.

[Halmyre]

witko999 said:

Regards the congestion, driverless cars are not going to magically make huge road capacity increases.

They may actually lead to an increase in congestion, since the inability to drive will no longer be a barrier to 'driving'.

Bad driving contributes little to the congestion on my daily commute. It's mostly down to the fact that a large number of people want to get to a limited amount of space at roughly the same time.

Getting away from the thread a bit, if you want to reduce congestion, encourage companies who employ people to sit in front of computer screens all day to implement a 'work from home' scheme. Reduces congestion, employee saves an hour or so on their commute and child-minding costs, employer saves providing office space.

[warp9]

There are lots of people on this thread that have quite a myopic view of whats happening in the larger world based around them being a car enthusiast.

There is massive investment and development like we have never had before into this sector. Industry and governments have a vision and time scale for this to happen, the lane assist technology is only an early start point which is already established into mainstream cars. This will only continue, albeit true driverless cars are at least IMO 20 years away.

While the general public as it stands (PH is not representative of general public) is cautious of this technology, acceptance and trust will increasingly be accepted particularly from the next generation of teenagers who already do have a different view of their relationships with technology and cars.

I also think there is a massive demand for this technology. 99% of journeys are done for function. There is nothing pleasant about driving in day to day rush hour traffic. With time and acceptance the average Joe will definitely go for this.

The insurance issue is already being addressed within the industry. Government have already made legislative changes to accommodate driverless cars. IMO all that will happen is the burden of liability will shift to the driverless car system provider and as a user you will pay your proportion of the cost for providing that liability. It will be like paying your share on a fleet policy.

So guys on PH: This is really going to happen. It will take a number of years to develop and refine systems and for the public to gain acceptance. It won't be as difficult as you think.

[youngsyr]

deckster said:

witko999 said:

Of course, none of these problems are insurmountable, but to think it's coming anytime soon is what's naïve.

Of course this is nub of the matter - what do we mean by 'soon'? I absolutely agree that it's not going to happen this year, or this decade. By 2030? Possibly. In my lifetime (being in my early 40s)? Very likely. In my kid's lifetimes? Without a doubt.

I will make a firm prediction now that that my children's generation will be the last to own driving licenses.

I think the level 3 autonomous cars (capable of driving themselves in all but emergency situations) will be around within 10 years, perhaps even 5 as they're not far off now. How long it takes to make that last step to completely remove the driver is anyone's guess as it's not just down to the technology (which I believe will be available within 10 years).

A lot will be down to putting the support systems in place, e.g. a database of roadworks and diversions to inform autonomous cars of temporarily changed road layouts.

Thankfully Google and several car manufacturers seem to be fully engaged in the process, so perhaps the arrival of the truly autonomous car will arrive at the shorter end of our expectations - I certainly hope so.

[EnglishTony]

otolith said:

I really don't see why this is complicated by autonomy - you insure the car against third party risks resulting from the use of it on the road. It makes no difference whether someone is manually driving it or not. It's not about who is at fault, it is about who is liable.

Liability is the point. Is the person in the vehicle liable or is it the person who programmed the vehicle's software?

Lower premiums you say? What and reduce the turnover of your company? Unlikely.

As for trusting any Govt, anywhere, to organise anything, er no.

[youngsyr]

It's probably also worth mentioning that a lot of this technology isn't new - my Audi was first registered in early 2007 (nearly a decade ago!) and can happily maintain a safe gap between itself and the car in front without any throttle/braking input from the driver, even if the driver in front decides to do an emergency stop.

It's particularly impressive when going around sweeping bends and you realise that the Audi is tracking a car that isn't even directly in front of it anymore.

[Emeye]

FredClogs said:

Ari said:

Emeye said:

I was the passenger in a brand new uncrashable XC90 that crashed into an Audi. The crash sensor failed as it was dirty, though the car had been cleaned a couple of days earlier. If the sensor had not failed the car could have taken over and stopped.

We are definitely a long way off being able to trust a self driving vehicle.

And the driver just sat there and watched it happen?

That's not a true story, all adaptive cruise control systems use radar, they're not "sensors" like reverse park sensors - a bit of dirt would make no difference, unless the dirt was a lump of lead.

Who made you the expert? I can only report what I experienced, as I was sat in the passenger seat of the car. Why the hell would I make it up?

Driver error was involved - an Audi turned right across us as we were attempting to turn right out of a junction. Due to the height of the Volvo bumper the black Audi was unsighted - a car behind the Audi had flashed us to drive out. The xc90 driver, my work buddy, immediately admitted responsibility,

The sensors should have saved us - I was a passenger in the same Volvo when some local Cornwall numpty with failed brake lights did an emergency stop in front of us at low speed and the xc90 reacted before my mate even had a chance - but it appears that the entire system had shut itself down on this occasion and was in error mode due to a single sensor error.

The driver had noticed something was wrong when we reversed as the parking assistant did not appear to be working.

After exchanging details, we tried restarting the car to see if the sensor system restarted and then noticed the error on the dash. We tried getting out the car and fully resetting the car. Still the error. I had a look at the sensors and notice the central one on the bonnet that looks like a camera was dirty.

I cleaned this sensor with my finger, we restarted the car and the system was working again. It has been ok since - the damage to the XC90 was just scuffs to the bumper and grill. The Audi had a huge bump in it.

So it seems on this occasion, as one of the sensors was in error, the whole system had shut down. Driver error had caused the crash, But it proves you cannot currently rely on technology to keep you safe.

ETA I have been told that an aluminium Crisp packet can seriously confuse VW's ACC.

Edited by Emeye on Thursday 2nd June 11:58

[otolith]

EnglishTony said:

otolith said:

I really don't see why this is complicated by autonomy - you insure the car against third party risks resulting from the use of it on the road. It makes no difference whether someone is manually driving it or not. It's not about who is at fault, it is about who is liable.

Liability is the point. Is the person in the vehicle liable or is it the person who programmed the vehicle's software?

Lower premiums you say? What and reduce the turnover of your company? Unlikely.

As for trusting any Govt, anywhere, to organise anything, er no.

The person liable is the one choosing to use the vehicle on the road. Same as they are liable if a mechanical failure causes the unattended car to roll down a hill and take out a bus queue.

[rxe]

heebeegeetee said:

rxe said:

but the "driver" has to be able to handle any situation, just as pretty much every human with a driving license can do.

There isn't much in your post that I wish to address, because I think all those issues will be addressed, and possibly already are apart from human frailty.

The only bit I fundamentally disagree is the quote above. I've driven a lot of miles in a variety of types of vehicles, and my opinion is that the average driver can not cope properly with any challenge at all. The vast majoroty of drivers drive the same car at the same times on the same roads and the same routes almost all of the time, and the moment they don't, such as weekends but especially holiday times, the casualty rates ramp right up.

My point is not that the automation can never make mistakes - but it does need to be able to work through any problem presented to it.

By way of example, how about low sun reflecting off a shiny road surface. The optical camera is saying "there something there, but I can't see it". The LIDAR is saying "nothing there". At that point the car is getting inconsistent data, and stops. It either demands a take over (which will be a problem if the driver is not in the car at the time), or just waits. When several thousand these cars do the same thing, we have a big problem.

Now the human will know what they are seeing, hopefully just slow down and drive through it. The other big problem is an object in the road. Every day we drive through litter, and make instant judgements about whether that bag has to be avoided (looks full) or can be driven over (is empty). How is machine vision meant to discriminate between a plastic bag blowing down the motorway, and something solid falling off the lorry in front? Both the optical camera and LIDAR are saying "solid object moving towards me at 10m/sec", and the car would have to take sudden avoiding action.

These are just two silly examples - pretty much on every journey, I see something strange that would defy a rules based system. If you look at the data coming out of the Google trials, you'll see that these are real problems that have not been solved yet.

I don't see insurance as a problem, that's just money and the will to do it. If the ability to drive home from the pub after a good night out costs me an increased premium, so be it, much better than getting cabs.

As someone else has said - this is all a question of timescale. I don't think this will be solved in a decade. Probably not for two decades. But eventually it will be solved.

[technodup]

witko999 said:

We have the option of car pooling right now but it's seldom used. In the future, why would anyone want to wait for a car to show up, then sit with 3 strangers while the car takes them to 3 other peoples place of work before finally arriving at their own?

I don't see it working like that at all (unless you choose it for cost purposes). I see it as individual cars, or pods or whatever picking me up, dropping me off and going to pick up someone else.

And as for car clubs two things are certain 1) they're much more in use now than ten years ago and 2) councils actively encourage their use via preferential spaces etc.

The change is already happening, it's just that some can't see it. The only topic of debate (well, anywhere else!), is the timescale.

[otolith]

I'm not really convinced that random posters on PH can think of problems that engineers employed building these systems cannot.

I also suspect that the things amateurs think are difficult probably aren't the problematic areas.

[robinessex]

My present car has developed and electrical fault, where all the warning lights go on and off at random. My son-in-laws car would stop randomly, and then re-start after a few attempts. My mates car does the same. They're not all that old. The garage man has confessed that these sorts of faults are common, and it results in what is a perfectly useable vehicle, with many more miles under it’s belt being scrapped as it’s just a gamble how long it can take to eventually (maybe) discover the fault, and that would cost more than the cars value. So what happens when these driverless cars do the same, at speed, on a motorway ? Even if the driver takes command, the faulty electronics may not allow any response by the car to driver inputs. Not for me I’m afraid.

[rxe]

otolith said:

I'm not really convinced that random posters on PH can think of problems that engineers employed building these systems cannot.

I also suspect that the things amateurs think are difficult probably aren't the problematic areas.

The problems I've mentioned are real ones that are being worked on by Google. They're hoping to get them fixed by 2020, and they will probably achieve it. The point is that anyone thinking this stuff is a few years away is delusional.

[Coolbanana]

warp9 said:

There are lots of people on this thread that have quite a myopic view of whats happening in the larger world based around them being a car enthusiast.

There is massive investment and development like we have never had before into this sector. Industry and governments have a vision and time scale for this to happen, the lane assist technology is only an early start point which is already established into mainstream cars. This will only continue, albeit true driverless cars are at least IMO 20 years away.

While the general public as it stands (PH is not representative of general public) is cautious of this technology, acceptance and trust will increasingly be accepted particularly from the next generation of teenagers who already do have a different view of their relationships with technology and cars.

I also think there is a massive demand for this technology. 99% of journeys are done for function. There is nothing pleasant about driving in day to day rush hour traffic. With time and acceptance the average Joe will definitely go for this.

The insurance issue is already being addressed within the industry. Government have already made legislative changes to accommodate driverless cars. IMO all that will happen is the burden of liability will shift to the driverless car system provider and as a user you will pay your proportion of the cost for providing that liability. It will be like paying your share on a fleet policy.

So guys on PH: This is really going to happen. It will take a number of years to develop and refine systems and for the public to gain acceptance. It won't be as difficult as you think.

Absolutely spot on!

The 'PH Expert's' who claim this isn't going to happen or is a long way off simply aren't the 'Expert's' that they think they are. Well, perhaps expert's at Pub talk and armchair criticism about things they have no real expert knowledge of.

Still, it is an interesting topic worthy of conjecture and discussion as it is sure to raise fears on many different levels amongst those who are technophobes or reluctant to accept change; they will undoubtedly want to see reasons for it to fail.

But I do believe it is now inevitable. Driverless cars are coming. A very good thing, in my own opinion. The commute should benefit hugely for all. Gone will be Mr Angry (Grr Grr Grr) and Mr Slowpoke in the Rush Hour. Instead, we will have cars 'speaking' to one another and ensuring far more efficient travelling for all.

Pistonheaders will still have their V8's tucked away for spirited runs too, no doubt. Driverless cars will not replace ALL cars albeit human-powered may find restrictions to certain roads in the future so as not to interrupt the Driverless journey's.

[technodup]

robinessex said:

Not for me I’m afraid.

My Dad said the same about flying, mobile phones, buying online, banking online and he says the same about this. He changed his mind on all of them, and if time allows will change his mind on this too.

Government has reduced smoking by about 3/4 by increasing the price and highlighting the dangers. Likewise seat belts, drink driving and now speeding. 'Normal' driving will come under the same pressure. People will do as they're told. Those who don't are punished by law.

The drip drip is already underway. Endless articles, news pieces and stats. Time builds trust, and there's plenty of time.

[Johnnytheboy]

technodup said:

Government has reduced smoking by about 3/4 by increasing the price and highlighting the dangers. Likewise seat belts, drink driving and now speeding. 'Normal' driving will come under the same pressure. People will do as they're told. Those who don't are punished by law.

Very true, we'll probably taxed into the damn things.

[boxedin]

Edinburger said:

The Docklands Light Railway is a driverless and successful railway.

I've test drove the Tesla Model S and used auto pilot and it's phenomenal. Definitely the way forward.

The DLR ( I think it was a Dutch company ) took many years to get working properly, needed lots of manual overrides. When it first went in it ran until 2100, not at weekends and only ran from Tower Gateway to Island Gardens ( I never ventured 'North' on it ). It was a very slow and long journey, and I don't just mean the physical speed it travels at! At least back then getting a seat was easy.

Having said all that, the Tesla blows the DLR into the weeds.

[youngsyr]

rxe said:

heebeegeetee said:

rxe said:

but the "driver" has to be able to handle any situation, just as pretty much every human with a driving license can do.

There isn't much in your post that I wish to address, because I think all those issues will be addressed, and possibly already are apart from human frailty.

The only bit I fundamentally disagree is the quote above. I've driven a lot of miles in a variety of types of vehicles, and my opinion is that the average driver can not cope properly with any challenge at all. The vast majoroty of drivers drive the same car at the same times on the same roads and the same routes almost all of the time, and the moment they don't, such as weekends but especially holiday times, the casualty rates ramp right up.

My point is not that the automation can never make mistakes - but it does need to be able to work through any problem presented to it.

By way of example, how about low sun reflecting off a shiny road surface. The optical camera is saying "there something there, but I can't see it". The LIDAR is saying "nothing there". At that point the car is getting inconsistent data, and stops. It either demands a take over (which will be a problem if the driver is not in the car at the time), or just waits. When several thousand these cars do the same thing, we have a big problem.

Now the human will know what they are seeing, hopefully just slow down and drive through it. The other big problem is an object in the road. Every day we drive through litter, and make instant judgements about whether that bag has to be avoided (looks full) or can be driven over (is empty). How is machine vision meant to discriminate between a plastic bag blowing down the motorway, and something solid falling off the lorry in front? Both the optical camera and LIDAR are saying "solid object moving towards me at 10m/sec", and the car would have to take sudden avoiding action.

These are just two silly examples - pretty much on every journey, I see something strange that would defy a rules based system. If you look at the data coming out of the Google trials, you'll see that these are real problems that have not been solved yet.

I don't see insurance as a problem, that's just money and the will to do it. If the ability to drive home from the pub after a good night out costs me an increased premium, so be it, much better than getting cabs.

As someone else has said - this is all a question of timescale. I don't think this will be solved in a decade. Probably not for two decades. But eventually it will be solved.

Your two examples (and many others besides) are easily answered.

Firstly, you're confusing the purposes of the radar and optical systems. Radar is used to judge objects, speeds and distances. Optical is used to judge road markings. In your example the car would be using the radar to judge whether an object is present, not the optical. There would therefore be no issue with sun glare or other optical illusions.

The empty bag example is also easily solved - radar is better at determining whether objects are solid or empty than the human eye, so the car would know for sure whether an object in front of it poses a risk, better than a human could and if the object does pose a risk, the car would take any safe avoiding action possible and do it much faster than a human could.