? It seems almost criminal to allow automated vehicles on the road as they can never be 100% reliable
Well we let humans out on the road, none of which are 100% reliable !
90% of road crashes are caused by human error so we have set the bar very low for automated vehicles
People won't accept automated cars that are not 99.9% reliable though
|
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
1500 x 2 million is 3 billion, not 3 trillion, a thousand times less.
Also, it would be interesting to see a detailed breakdown of the quoted £2 million cost.
Also a deceased person does not draw a State pension, which is a small offsetting saving, won't require further NHS services or other benefits . (heartless, but if calculating costs of a fatality these sums are relevant)
Edited by galileo on 04/06/2020 at 18:09
|
|
Fully autonomous vehicles that share normal roads, and are not guided by a buried wire or a remote signal will never happen until it is possible to persuade an insurance underwriter to insure it with no get out of jail weasel words in the contract. I don't see a queue of them forming yet!
|
Fully autonomous vehicles that share normal roads, and are not guided by a buried wire or a remote signal will never happen until it is possible to persuade an insurance underwriter to insure it with no get out of jail weasel words in the contract. I don't see a queue of them forming yet!
Widespread adoption of fully autonomous road vehicles is still a long way off but it will happen eventually. The legal and insurance problems will be bypassed by manufacturers retaining ownership of the vehicles and being responsible for maintaining them.
|
|
But manufacturers won't assume liability for their current vehicles, never mind future autonomous ones! The dealer takes the can back at the end of the day.
No manufacturer is going to assume infinite public liability for it's products in the UK.
Like I said - it's all down to underwriters accepting the risk.
|
|
|
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
1500 x 2 million is 3 billion, not 3 trillion, a thousand times less.
Also, it would be interesting to see a detailed breakdown of the quoted £2 million cost.
Also a deceased person does not draw a State pension, which is a small offsetting saving, won't require further NHS services or other benefits . (heartless, but if calculating costs of a fatality these sums are relevant)
Sorry about the maths, you are correct £3Billion
Cost stated at £2.2M
www.statista.com/statistics/322862/average-cost-of.../
Looks like this does not include Insurance payouts to relatives. For example a high earner with 3 young children, payout would be much more than a couple of million.
Also Personal injury, what if person above needed lifelong care as well.
Driverless cars will be a reality because us as humans are very bad at driving. Fact not opinion.
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
1500 x 2 million is 3 billion, not 3 trillion, a thousand times less.
Also, it would be interesting to see a detailed breakdown of the quoted £2 million cost.
Also a deceased person does not draw a State pension, which is a small offsetting saving, won't require further NHS services or other benefits . (heartless, but if calculating costs of a fatality these sums are relevant)
Sorry about the maths, you are correct £3Billion
Cost stated at £2.2M
www.statista.com/statistics/322862/average-cost-of.../
Looks like this does not include Insurance payouts to relatives. For example a high earner with 3 young children, payout would be much more than a couple of million.
Also Personal injury, what if person above needed lifelong care as well.
Driverless cars will be a reality because us as humans are very bad at driving. Fact not opinion.
The £2.2 M in the statista link is actually listed as "prevention costs", so is this the cost of road safety actions to prevent fatalities?
I would think that a non-fatal accident with life - changing imjuries would cost the country more than a fatal one, considering treatment and long term support for the injured party, insurance costs could be comparable too?
Edited by galileo on 05/06/2020 at 09:34
|
|
|
|
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
It will come but will be a lot longer than 3/5 years that a lot of the tech companies are saying it will.
|
|
Totally agree - it's not a matter of if but when.
Human beings are hugely flawed - they drive tired, angry, emotionally, hungry, thirsty, needing a pee etc etc. It's no wonder they have so many "accidents".
I suspect that autonomous vehicles will come sooner than we think. Possibly not on high density, idiosyncratic UK roads, but overseas where roads are less congested, newer and legalities more easily solved. This will allow real operational gremlins to be resolved in a way that controlled testing can't.
The social and economic benefits of autonomous vehicles are overwhelming. If there are accidents the "black box" will be able to identify precisely what went wrong, unlike distorted human recollections concerned about saving face, passing the blame, preserving their NCB.
|
|
Motor business likes flawed. They dont like taking the responsibility for software - Dieselgate being the obvious example.
Safety devices, such as city braking, lane keep assist, are not autonomous. And lane keep assist is still fallible - ask a BMW driver who's tried to cross a centre road line to pass a tractor or cyclist - the wheel fights to stay in lane.
I guarantee autonomous, autopilot cars will not be here soon, and will likely not be worth the investment before too long.
|
|
|
|
Except people don't make the decisions authorities do.
If automation reduces the fatalities on roads by 90% then about 1500 people less will die on the UK roads. If average cost of 1 death is £2 million that's £3 trillion in UK alone. Extrapulate that worldwide figure is massive.
Unless us humans seriously improve our game automation is coming, like it or not.
It will come but will be a lot longer than 3/5 years that a lot of the tech companies are saying it will.
That will depend on the hardware needing less power and adaptable software, it is said the vehicles at the moment are supercomputers that need reducing in size and power.
ie power requirements are high for the sensors and computers which competes with the drivetrain for battery power. this will explain https://www.reuters.com/article/us-tech-ces-qualcomm/qualcomm-launches-autonomous-driving-computer-aiming-to-hit-roads-by-2023-idUSKBN1Z51YH
which looks good and interesting to see if they meet targets
|
|
A laptop consumes about 50 watts per hour.
An EV consumes about 15 kwh to travel around 40 miles in an hour.
So even if my EV uses the equivalent (say) 3 laptops to run the sensors and computers, this would total 150 w in an hour - about 1% of the total power usage. I can surf the web on a basic smartphone for a couple of hours using only a tiny battery
I think there is more than a little hype in the Reuters article!!
|
|
The computers that control self driving cars are very powerful. Multicore high end graphics chips struggle to keep up with the demands put on them. Graphics chips are huge consumers of power. Think kilowatts not watts with enough heat to boil water and significant active cooling is required. Add lidar, radar, camera and ultrasonics and it's not unexpected to see the increased power requirements.
|
The computers that control self driving cars are very powerful. Multicore high end graphics chips struggle to keep up with the demands put on them. Graphics chips are huge consumers of power. Think kilowatts not watts with enough heat to boil water and significant active cooling is required. Add lidar, radar, camera and ultrasonics and it's not unexpected to see the increased power requirements.
Will they have much in the way of graphics chips for what they do? Gaming PC's usually use around 500/600w power supplies.
|
The computers that control self driving cars are very powerful. Multicore high end graphics chips struggle to keep up with the demands put on them. Graphics chips are huge consumers of power. Think kilowatts not watts with enough heat to boil water and significant active cooling is required. Add lidar, radar, camera and ultrasonics and it's not unexpected to see the increased power requirements.
Will they have much in the way of graphics chips for what they do? Gaming PC's usually use around 500/600w power supplies.
more than that my pc with near basic Nvidia card has 750w power supply, some gaming pcs use in excess of 1500 watt supplies which is easy to check out. and cooling is usually water cooled on gaming machines to keep up with cooling
so imagine several chips using all resources to control a vehicle and that will need serious cooling, as when they get too hot they throttle down and do not work as well which you wont want on a vehicle
|
|
|
The computers that control self driving cars are very powerful. Multicore high end graphics chips struggle to keep up with the demands put on them. Graphics chips are huge consumers of power. Think kilowatts not watts with enough heat to boil water and significant active cooling is required. Add lidar, radar, camera and ultrasonics and it's not unexpected to see the increased power requirements.
Will they have much in the way of graphics chips for what they do? Gaming PC's usually use around 500/600w power supplies.
Funnily enough yes. Not to display graphics though!
Graphics cards are designed to shift data. Copying huge chunks of data from one memory location to the next and applying matrix (x,y,z grid) manipulation of the data. Matrix manipulation is very important in AI but I think this is more about placing a scanned object in the computers memory and overlaying other data.
The graphics card companies had been selling custom versions of their cards with thousands of cores to the car manufacturers but now they are using their expertise to design off the shelf automotive systems for the car manufacturers to provide astonishing computing power at much lower power requirements.
Edited by Zippy123 on 06/06/2020 at 00:53
|
|
The graphics card companies had been selling custom versions of their cards with thousands of cores to the car manufacturers but now they are using their expertise to design off the shelf automotive systems for the car manufacturers to provide astonishing computing power at much lower power requirements.
bearing in mind that graphics cards are built into processors now which though not so good for gaming, will be good enough for a car display, in fact most parts of a pc are now in a chip
|
bearing in mind that graphics cards are built into processors now which though not so good for gaming, will be good enough for a car display, in fact most parts of a pc are now in a chip
I think that you have missed the point. The graphics processors are not used for displaying graphics in cars but for shifting huge quantities of data captured by the multitude of onboard sensors required in self driving cars.
|
bearing in mind that graphics cards are built into processors now which though not so good for gaming, will be good enough for a car display, in fact most parts of a pc are now in a chip
I think that you have missed the point. The graphics processors are not used for displaying graphics in cars but for shifting huge quantities of data captured by the multitude of onboard sensors required in self driving cars.
Think he covered it in the first sentence 'but now they are using their expertise to design off the shelf automotive systems for the car manufacturers to provide astonishing computing power at much lower power requirements.'
|
bearing in mind that graphics cards are built into processors now which though not so good for gaming, will be good enough for a car display, in fact most parts of a pc are now in a chip
I think that you have missed the point. The graphics processors are not used for displaying graphics in cars but for shifting huge quantities of data captured by the multitude of onboard sensors required in self driving cars.
Think he covered it in the first sentence 'but now they are using their expertise to design off the shelf automotive systems for the car manufacturers to provide astonishing computing power at much lower power requirements.'
The first sentence was the last sentence from my previous post. The quote marker was not used.
|
|
I think that you have missed the point. The graphics processors are not used for displaying graphics in cars but for shifting huge quantities of data captured by the multitude of onboard sensors required in self driving cars.
I didnt explain myself properly as usual so sorry about that, my point was meant to be that you can have all the processor power you want, the problem comes with the data getting to it and data being transferred to the car components, so unless the processor and memory have multiple/quad lanes-threads the processor could stall waiting for the info it needs which you dont want
dont forget it took years before CPUs got double or multiple threads because they cannot increase the clock speed, so they use multiple cores instead and now they are stacking them iirc, I think intel started years ago
the other obvious problem is getting the software right which will take longer than the hardware development imo.
|
|
|
|
|
|
|
