Electric cars

Car_54

NigeWick wrote: »

But then just over a century ago, the overwhelming majority thought powered transport would never catch on

Despite the existence of a nationwide railway network, and the replacement of sail by steam?

zeupater

NigeWick wrote: »

I believe by 2021, yes. And, by 2030 all new vehicles will be autonomous electrically powered ones.

I readily accept that at present the overwhelming majority will think that's impossible. But then just over a century ago, the overwhelming majority thought powered transport would never catch on and that heavier than air flight was impossible too.

Hi

The issue is that there are effectively three hurdles ...

(1) ... Test & Approval Timeline
(2) ... Cost of AI processing & vehicle automation
(3) ... Consumer Acceptance

Firstly, the testing cycle needs to demonstrate that autonomous vehicles are inherently safer than current manual ones. The USA seems to be at the forefront of development & testing and the recent spate off accidents doesn't bode too well even though most roads are straight, wide & not congested, so nothing like European conditions.

The latest fatal accident involving a pedestrian will have serious implications on the approval timeline as any chance of approval will depend on a simple metric - fatalities per million miles driven ... currently within the US it's somewhere around 1:100million, so in open road testing the mileage bar in the USA has just been raised by something like a factor of 10 before approval would even be considered, which seriously impacts either the testing timeline or costs involved (ie 10x longer or 10x the vehicles!) ... it's a standard project management balancing approach.

On cost ... the processing power required for autonomous vehicles is huge, somewhere approaching that of 100 laptops/car & that kind of processing isn't cheap and won't be for 8-10 years - add in the mechanics for converting processing power into a viable vehicle & the on-cost becomes a considerable burden for years to come, disproportionately so as the model aspirational element and relative price falls ... I'd certainly expect luxury marques to be running AI for at least 10 years before the technology enters the 'everyday' volume production sector which makes-up the vast majority of what's on the roads ... think about how long it's taken for safety related developments such as ABS to move from high-end vehicles to mid-market models and the technology there is relatively simple.

On acceptance ... Consumers are pretty slow to take up technology they don't understand, don't want, don't see a need for, or have any issues with ... look at smart-meters, ten years ago there was an aspirational requirement that we were all to have them installed by now, yet take-up is remarkably low and currently (when most should already have been installed) somewhere around 1 in 4 people consulted think that they'll provide any benefit ... and that's without having to consider paying thousands extra for the option either 'up-front' or within lease agreements!

In 2030 the vast majority of vehicle on the road will either be current models or variations on those models and the overwhelming proportion of new vehicle build & sales will be manually driven ... it's almost certain - the combination of timeline, cost, technical learning curve and consumer viewpoint simply can't be ignored no matter how it looks to those looking through rose tinted spectacles.

HTH
Z

Martyn1981

zeupater wrote: »

Hi

The issue is that there are effectively three hurdles ...

(1) ... Test & Approval Timeline
(2) ... Cost of AI processing & vehicle automation
(3) ... Consumer Acceptance

Firstly, the testing cycle needs to demonstrate that autonomous vehicles are inherently safer than current manual ones. The USA seems to be at the forefront of development & testing and the recent spate off accidents doesn't bode too well even though most roads are straight, wide & not congested, so nothing like European conditions.

The latest fatal accident involving a pedestrian will have serious implications on the approval timeline as any chance of approval will depend on a simple metric - fatalities per million miles driven ... currently within the US it's somewhere around 1:100million, so in open road testing the mileage bar in the USA has just been raised by something like a factor of 10 before approval would even be considered, which seriously impacts either the testing timeline or costs involved (ie 10x longer or 10x the vehicles!) ... it's a standard project management balancing approach.

On cost ... the processing power required for autonomous vehicles is huge, somewhere approaching that of 100 laptops/car & that kind of processing isn't cheap and won't be for 8-10 years - add in the mechanics for converting processing power into a viable vehicle & the on-cost becomes a considerable burden for years to come, disproportionately so as the model aspirational element and relative price falls ... I'd certainly expect luxury marques to be running AI for at least 10 years before the technology enters the 'everyday' volume production sector which makes-up the vast majority of what's on the roads ... think about how long it's taken for safety related developments such as ABS to move from high-end vehicles to mid-market models and the technology there is relatively simple.

On acceptance ... Consumers are pretty slow to take up technology they don't understand, don't want, don't see a need for, or have any issues with ... look at smart-meters, ten years ago there was an aspirational requirement that we were all to have them installed by now, yet take-up is remarkably low and currently (when most should already have been installed) somewhere around 1 in 4 people consulted think that they'll provide any benefit ... and that's without having to consider paying thousands extra for the option either 'up-front' or within lease agreements!

In 2030 the vast majority of vehicle on the road will either be current models or variations on those models and the overwhelming proportion of new vehicle build & sales will be manually driven ... it's almost certain - the combination of timeline, cost, technical learning curve and consumer viewpoint simply can't be ignored no matter how it looks to those looking through rose tinted spectacles.

HTH
Z

Hiya. If Tony Seba is correct then for processing these cars need about 8 teraflops, and I think the Model 3 has 10teraflops, apparently the Xbox one is 6.

For trialing, that'll take a lot of practice, yes, but it's happening already, even with some cars that have no steering wheel. So progress could be dramatically fast - possibly just a few years, maybe early 2020's.

Then there's the chance that cars will have the ability to be guided by a base controller in very strange situations where the computer says 'no'.

For acceptance, I really don't think this will be slow. Non drivers, especially the young will jump on the opportunity of super cheap taxi's, and the rest of us will follow when we realise how much money we're wasting on a car that sits outside 95% of the time.

Even if acceptance is a little slow, I suspect that 2nd, 3rd, 4th etc cars will disappear very fast.

I think the delay will be supply, not demand, but, having been caught out so many times with my optimistic guesses for PV and wind being massively behind the actual rollout and cost reductions, I'm tempted to say that the companies will switch over faster than we expect.

If EV's are now the cheapest option already, when all costs are taken into account, then a customer demand switch could be dramatic as prices continue to fall fast over the next few years.

So I'm confident of a fast changeover ..... but ..... I don't think the big automotive companies are truly on board yet, so it may take longer simply as they drag their heels until the point they realise they could lose massive turnover forever to the new kids on the block.

Please bear in mind that I'm openly guessing, and basing expectations on a massive demand led change, but I don't really know which way it'll go.

PS - back to two hands!

zeupater

Martyn1981 wrote: »

Hiya. If Tony Seba is correct then for processing these cars need about 8 teraflops, and I think the Model 3 has 10teraflops, apparently the Xbox one is 6.

For trialing, that'll take a lot of practice, yes, but it's happening already, even with some cars that have no steering wheel. So progress could be dramatically fast - possibly just a few years, maybe early 2020's.

Then there's the chance that cars will have the ability to be guided by a base controller in very strange situations where the computer says 'no'.

For acceptance, I really don't think this will be slow. Non drivers, especially the young will jump on the opportunity of super cheap taxi's, and the rest of us will follow when we realise how much money we're wasting on a car that sits outside 95% of the time.

Even if acceptance is a little slow, I suspect that 2nd, 3rd, 4th etc cars will disappear very fast.

I think the delay will be supply, not demand, but, having been caught out so many times with my optimistic guesses for PV and wind being massively behind the actual rollout and cost reductions, I'm tempted to say that the companies will switch over faster than we expect.

If EV's are now the cheapest option already, when all costs are taken into account, then a customer demand switch could be dramatic as prices continue to fall fast over the next few years.

So I'm confident of a fast changeover ..... but ..... I don't think the big automotive companies are truly on board yet, so it may take longer simply as they drag their heels until the point they realise they could lose massive turnover forever to the new kids on the block.

Please bear in mind that I'm openly guessing, and basing expectations on a massive demand led change, but I don't really know which way it'll go.

PS - back to two hands!

Hi

Possibly, but the question is whether the Model 3 is fully autonomous or simply has the ability to lane follow, park and do some basic intelligent functions as opposed to being fully autonomous ... after-all, depending on how critically you apply the definition, it has level 2 autonomy, possibly bordering on level 3.

From what I've seen on the technology side you're really talking about going north of 300trillion operations/second, so at least 30x the processing power you mention .... this is a major stumbling block considering that we're talking CPU processing as opposed to the rendering abilities of a gaming machine GPU ... for example, Intel's (really expensive) i9 processor is rated at just about 1teraflop ....

Herein lies the issue, and it's massive .... EVs are here and ready to be built in volume, yet we're only just looking at moving from Level2 driving autonomy and we need to get to Level5, not only that, but an affordable Level5 which doesn't have too great an impact on the cost of entry level to mid range vehicles which comprise the vast majority of vehicle sales ....

HTH
Z

Martyn1981

zeupater wrote: »

Hi

From what I've seen on the technology side you're really talking about going north of 300trillion operations/second, so at least 30x the processing power you mention .... this is a major stumbling block considering that we're talking CPU processing as opposed to the rendering abilities of a gaming machine GPU ... for example, Intel's (really expensive) i9 processor is rated at just about 1teraflop ....

HTH
Z

Is it really that high? Maybe that's the issue, we're seeing different claims from different people. I thought the Model 3 had the computing power (but not the kit), so that side was sorted, but perhaps that was BS.

Just to be clear, I appreciate everything you are saying, but we are both giving opinions and assumptions, whereas reality, in these fields, can be way faster, or slower than expected. So really it's just a case of when the disruption starts, early 2020's, mid 2020's, late 2020's, but it will happen.

By co-incidence, this afternoon I watched a fun you-tube on this very issue published today, trying to think through the issues. You may or may not agree, but I think it's a pleasant watch. I found it because the presenter is part of a father and son team that do an interesting weekly video called Tesla Time News.

Will There be Problems in Our Autonomous Driving Future?

Martyn1981

bugslet wrote: »

We double man a lot of our vehicles and the driving time non-stop is 21 hours, though we probably only do that a couple of times a year. Those double manned Waberer trucks will be maxing the hours though:eek:

We have one run which takes 23 hours and we stage crews mid way along the run at hotels as we aren't allowed to stop. Other moves are limited to very specific staging places. So I'd be looking at a range of at least 800 miles

If I was in general haulage though, then around 400 miles would be OK. If I was a major haulier trunking between depots, then I'd have charging points on site and a range of around 200 miles or less would be adequate.

There's never a definitive answer in haulage!

I'm not a bulk carrier, but no I don't believe distance would be as critical, you don't see many bulk carriers heading in and out of Europe. They will be doing everything they can to minimise the weight however.

Hiya, can I ask some questions that you might be able to guess at, just for fun.

What percentage of the UK market do you think a 300 mile range truck with no payload penalty, or a 500 mile range truck with a 1tonne payload penalty, might be suitable for (each or separate)?

I'm simply fascinated by the possibilities. I'm blindly guessing at perhaps 50%+, but assume that even 10% of the market would be a massive gain for Tesla.

Of course that's assuming that their claims are true, and the extra cost of the Tesla can be made back in a year or so, given the high cost of diesel in the UK/EU.

I'm looking at this purely from numbers, not emotionally, as that factor may push the issue even harder due to real (or fake) green choices.

Thanks.

zeupater

Martyn1981 wrote: »

Is it really that high? Maybe that's the issue, we're seeing different claims from different people. I thought the Model 3 had the computing power (but not the kit), so that side was sorted, but perhaps that was BS.

Just to be clear, I appreciate everything you are saying, but we are both giving opinions and assumptions, whereas reality, in these fields, can be way faster, or slower than expected. So really it's just a case of when the disruption starts, early 2020's, mid 2020's, late 2020's, but it will happen ...

Hi

To simplify the issue, what needs to be considered is that the relative processing power of what's in control of the vehicle is a decent place to start .... 1 Trillion operations per second (teraflops) sounds huge, but the fastest supercomputers currently in operation have peak processing capabilities well over 1000 times as fast (a petaflop), some of the latest proposals 200,000 times as fast, that's approaching 0.2 of an exaflop which is in the region of what a human brain is reckoned to process short-term memory at .... in short, the lump of meat in your head is likely to have peak processing power of 1million Intel i9 processors, 100,000 Model 3 cars, 3000x more than that estimated for Level5 autonomous vehicles, or 5 times that of the fastest supercomputers.

Now, do we really need all of that power (1/3000th of that of a human brain)? ... to run code to control a vehicle in a controlled environment probably not, but what we're taking about is AI - the ability to recognise both environment & situation then continually take appropriate actions within as short a timescale as possible even if the situation hasn't been encountered before based on complex algorithms that replace something which we all take for granted, intuition - after-all lives are at risk, and intuitive comparison to experience takes far more processing power than conscious reasoning or execution of hard written code ...

Don't misunderstand what I'm saying, I believe that fully autonomous vehicles will happen eventually, but Moore's law is starting to slow down with additional power now looking at requiring additional processors and those high end processors consume high-end budgets and bucket-loads of energy ... that's why the level of autonomy we'll see in most vehicles is likely to be Level2 or Level3 for a considerable time yet, so a driver aid more than a fully automated 'whistle and one will come & get you wherever you are' solution that some believe will be available within just a few months .... and that's why the ownership model and most vehicles on the road in 2030 will look very similar to what they do today, other than maybe somewhere around 1 in 6 cars on UK roads being 'electrified' to some degree or other!

HTH
Z

Martyn1981

zeupater wrote: »

Hi

To simplify the issue, what needs to be considered is that the relative processing power of what's in control of the vehicle is a decent place to start .... 1 Trillion operations per second (teraflops) sounds huge, but the fastest supercomputers currently in operation have peak processing capabilities well over 1000 times as fast (a petaflop), some of the latest proposals 200,000 times as fast, that's approaching 0.2 of an exaflop which is in the region of what a human brain is reckoned to process short-term memory at .... in short, the lump of meat in your head is likely to have peak processing power of 1million Intel i9 processors, 100,000 Model 3 cars, 3000x more than that estimated for Level5 autonomous vehicles, or 5 times that of the fastest supercomputers.

Now, do we really need all of that power (1/3000th of that of a human brain)? ... to run code to control a vehicle in a controlled environment probably not, but what we're taking about is AI - the ability to recognise both environment & situation then continually take appropriate actions within as short a timescale as possible even if the situation hasn't been encountered before based on complex algorithms that replace something which we all take for granted, intuition - after-all lives are at risk, and intuitive comparison to experience takes far more processing power than conscious reasoning or execution of hard written code ...

Don't misunderstand what I'm saying, I believe that fully autonomous vehicles will happen eventually, but Moore's law is starting to slow down with additional power now looking at requiring additional processors and those high end processors consume high-end budgets and bucket-loads of energy ... that's why the level of autonomy we'll see in most vehicles is likely to be Level2 or Level3 for a considerable time yet, so a driver aid more than a fully automated 'whistle and one will come & get you wherever you are' solution that some believe will be available within just a few months .... and that's why the ownership model and most vehicles on the road in 2030 will look very similar to what they do today, other than maybe somewhere around 1 in 6 cars on UK roads being 'electrified' to some degree or other!

HTH
Z

Hi Z, that all makes sense, but what I was saying is that the figure I've seen touted for fully autonomous is 8Tfps, and the M3 apparently has 10 of them.

So the issue of processing power for me, is whether or not that 8 figure is BS ... which it may of course be.

But if not, then we may have autonomous taxi's by the early 2020's.


Regarding private ownership and the cars on the road, I suppose it's all down to how we, as humans react. Even if supply is too low, will we still buy an ICE if the EV's prove to be far, far cheaper by, say 2020-22 (based on all ownership costs)? What I'm wondering is whether we will still buy a 'horse', just because there's a shortage of 'Model T's' IYSWIM, I really don't know what the reaction will be, or if the big boys will be able to design, build and deploy EV's faster somehow.

Edit - This article does not support my opinion that it'll happen fast. ;-(

BMW Pushes Back EV Mass Production Again, Says Not Profitable Enough Until 2020


Further edit, following a night's sleep - the issue that the BMW article raises for me is complex, does it suggest:-

1. That they will sit back, then start development around 2020, meaning another 5yrs or so for mass deployment, or

2. That they will produce a decent car before 2020, and then when costs fall, they will be ready to switch over production lines, duplicate what already exists and mass produce from 2020?

The two scenarios appear to have wildly different results (time wise), and I'm not optimistic that it's option 2.

AnotherJoe

If EVs are cheaper (or should i say when, I believe thats a given, may be 5 may be 10 years away but it will happen) that is an argument against autonomous cars being used in preference to purchase.

Eg lets say an EV is half the cost to own and run as a petrol. That makes the difference between buying an EV and using autonomous services smaller than going from petrol to autonomous, eg less likely.

The whole argument about how autonomous cars will supplant ownership misses multiple obvious issues.

Convenience. I can just get in my car and go right now i dont have to call for one and wait for it nor do i need to plan.

Convenience. The two child seats and buggy are already in my car, i dont need to wrestle with getting them in (and where the heck do i put them once the autonomous car has headed off to pick up its next fare? ). I'm taking them to swimming lessons and while they are there nipping to the supermarket then back to collect them than after that little Timmy is going to Harry's house for tea then i'm going back home. Or, my golf clubs, or overnight bag, whateve,r are in my car at work so i can just get on to my golf game or overnight stay with no hassle of removing it from car.

Cost. Little Timmy just threw up in the back of the autonomous car theres a £100 cleaning fee.

Unpleasantness. I've just got in a car someones kid puked in its vile I'm redirecting the car to go to the nearest Tesla dealer so i can buy one and not have to experience that again.

Unpleasantness. I've just had an asthma attack / got fleas because someones dog was in this car before me. Its vile I'm redirecting the car to go to the nearest Tesla dealer so i can buy one and not have to experience that again.


Real life stuff like this is why people wont replace their owned (or leased etc) cars with autonomous ones they are just too inconvenient and awkward.

If they were going to work, people would already use taxis instead of owning cars, but they dont. Autonomous cars just are cheaper taxis, theres nothing special about them.

People may well supplement usage of their cars with autonomous ones, just like when i go to London I take a train, i dont drive in or sometimes i get a taxi so i can drink. But the existence of the train or taxi doesn't mean i got rid of my car, nor would my local taxi service halving its fees mean i dump my car, even if they were free i woudlnt dump my car i'd just use them more often.

NigeWick

Car_54 wrote: »

Despite the existence of a nationwide railway network, and the replacement of sail by steam?

Aah, another founder member of Pedants R Us! :beer:

Motor powered road transport, AKA the horseless carriage.;)