Volvo to 'accept full liability' for crashes with its driverless cars Volvo will "accept full liability" for collisions involving its autonomous vehicles, the company has confirmed. The announcement was made by Volvo's president and chief executive Håkan Samuelsson during a speech in Washington on Thursday. Ben Gardner, expert in autonomous vehicles technology and regulation at Pinsent Masons, …
>My car is old and falling apart. I want a new one. Ram the driverless Volvo!
Their log files will prove your liability.
>What if a driverless Volvo and driverless Tesla hit each other?
They won't. Not hitting other vehicles is probably one of the least hard problems for autonomous vehicles. If they ever do then, again, log files.
>Ooo, look - I found the factory configuration menu of the neighbour's car. What if I change drive on left to drive on right?
Mwahahaha... THAT'S the problem. Hacking.
Not hitting other vehicles is probably one of the least hard problems for autonomous vehicles.
Because cars are simple machines that never suffer mechanical failure, and there are no external failure modes (road surface failures, objects suddenly intruding into the path of the vehicle) that could ever cause two autonomous vehicles to crash.
Why is it that even the autonomous-vehicle objectionists so often ignore the full range of failure modes? Technology solves problems right up to the point where it doesn't. AVs might greatly reduce the number of collisions, but they can't prevent them.
If they ever do then, again, log files.
Because there are no failure modes that wouldn't leave log files intact. Again, this won't cover all cases, so we still have to decide how to deal with the remaining ones.
My car is old and falling apart. I want a new one. Ram the driverless Volvo!
Then I look forward to seeing you on "The World's Craziest Drivers - It's Not Just Russia Anymore" compilation coming soon to Youtube.
As for hacking - if you have access and intent why not just cut the brake line. It's a more plausible method of causing mayhem than trying to make a car drive on the wrong side of the road.
I could be wrong, but I doubt that they are proposing to unconditionally accept fault for all accidents invloving one of their cars, that would be nuts; I would imagine that they are absolving the "driver" (not sure what to call the person in the wheel seat of an autonomous vehicle) of liability. Look at the paragraph
"Volvo wants to remove the uncertainty of who would be responsible in the event of a crash," Gardner said. "At the moment it could be the manufacturer of the technology, the driver, a maker of a component in a car."
I read that as talking aboput the case of a crash *caused by the Volvo*, and then the question is is it the driver, manufacturer or component maker who takes responsibility? If the crash was found to be caused by the other vehicle, it wouldn't be an issue, the other guy is responsible.
> Ram the driverless Volvo!
Are you aware that the driverless cars are fitted with cameras and lasers to scan the area around to them to enable them to *be* driverless? I think they'd be recording that data...
> What if a driverless Volvo and driverless Tesla hit each other?
Then one or both must have had a major systems failure (or been shunted into the other by another vehicle)
> I found the factory configuration menu of the neighbour's car.
Now that one *is* a serious concern, hopefully there's no hard-coded passwords or other such nonsense caused by Sales taking precedence over Security!
> hopefully there's no hard-coded passwords or other such nonsense caused by Sales taking precedence over Security!
Chances of these systems coming to market without exploitable security holes? They will exist and they will be found. Given that current connected vehicle systems don't seem to isolate the canbus from the infotainment systems, which are in turn connected to the mobile network, I can't see the bean counters allowing anywhere near enough money to be spent on R&D in this area - it'll be a case of 'we've gotta be first to market/we've gotta get to market asap now that our competitors are there, get it working and start selling 'em NOW'
"Ooo, look - I found the factory configuration menu of the neighbour's car. What if I change drive on left to drive on right?"
There's no such setting. You'd have to get every other car on the roads to drive on the wrong side to fool the autonomous car - and then it's not really like it's wrong to drive on that side of the road with them.
Even if there were such a setting, the car simply wouldn't move, what with sensing all the people apparently driving the wrong way down the street.
Go a bit further.. Here in the States, auto accidents are major source of revenue for law firms. Also traffic violations are a major source of revenue for local governments via the police traffic enforcement. I suspect that politics and greed are about to rear it's ugly head over this.
I believe doing this will be far more difficult with driverless cars. To begin with, the driverless car will be far more adept at avoiding the collision than a normal driver. And even if you manage it, the driverless car will have complete records of the accident, including the suicidal behavior which caused it.
An autonomous vehicle will react more quickly and more safely than you ever can. By the time you're brain has processed the fact that the brake lights have come on on the car in front an AV will already have started braking (and will have probably done a better job of maintaining a suitably safe distance between you and the car in front to begin with).
When Volvo say they'll accept 'full liability' I'm sure they don't mean to give a free meal to every idiot who deliberately rams a driverless car. They mean that they will not slow down liability payouts by trying to legally offload onto their own suppliers. If you do ram a driverless car, you can bet that there will be efforts by Volvo, and others, to establish the facts and causes of the incident and to establish blame, as is perfectly normal in all traffic incidents.
At this early stage of development, it would be fairly easy to have small video cameras with outside views making a 5 minute (or whatever) looped recording onto an on-board SSD, along with car speed parameters, etc. This would be designed stop recording in the event of a crash or collison and so provide some evidence as to what had actually happened.
Yep, that's right. In any accident the first thing that is established is who is at fault, then it's their insurance who have to pony up (less any excess that the driver/owner needs to pay, and excluding any other behaviour such as deliberate crash that might exclude insurance liability)
What Volvo are saying is that if the crash is caused by their car, the liability will not fall on the owner and their insurance, but will be taken up directly by Volvo. This means that it should be incredibly cheap for a private person to insure a driverless Volvo, and thus removes the biggest obstacle to takeup*
Kudos to Volvo for putting their money where their mouth is, hopefully Google etc will follow suit
*EVERY article ever written on El Reg about driverless cars, someone in the forum pops up with "who's going to pay for it if/when they crash"
EVERY article ever written on El Reg about driverless cars, someone in the forum pops up with "who's going to pay for it if/when they crash"
I do. And now we have a car company saying quite unambiguously that they accept the blame for faults in their car design or manufacture, and that is a great step forward (subject to getting country laws to accepts such a thing).
As other commentards have pointed out, an autonomous car will almost certainly out-brake a human driver in any obvious impact scenario. Though how well they will deal with odd cases, loss of communications (doh! stupid idea...) and anticipation of kids, etc, playing at the roadside is another more difficult question to be answered.
Finally, can we please have proper audits and standards for car software? It is shitty enough we have cars recalled due to potential hacking via in-car entertainment (e.g. Jeep) and not shutting off (e.g. Ford) but having full control of all aspects of the vehicle offers far more opportunities for a BSOD than so far (e.g. Toyota's "unintended acceleration").
"...Finally, can we please have proper audits and standards for car software? It is shitty enough we have cars recalled due to potential hacking via in-car entertainment (e.g. Jeep) and not shutting off (e.g. Ford) but having full control of all aspects of the vehicle offers far more opportunities for a BSOD than so far (e.g. Toyota's "unintended acceleration")..."
Actually I would be more concerned with the likes of Volkswagen. We are already in a position where they have duped testers and yet the board (an incredibly controlling, micromanaging one at that by all accounts), claim no one knew...that it was down to one or two engineers...
So in this regard, I agree that it needs to be open, transparent and audited.
Re: proper audits for car software.
The good thing about the car manufacturer accepting liability like this is that market forces are correctly aligned with the interests of consumers and there's no need for complex legislation to impose proper audits. The car manufacturer, or their insurer, will want to make sure that the software works because if it doesn't they'll end up paying for crashes. This is very different to the VW situation, where there's a bit of software that wasn't really in the interests of either the car owner or the car manufacturer; not unsurprisingly this software actually did what the owner and manufacturer would want it to. I'd also assume that an autonomous volvo won't leave the drive until it's checked for security updates; volvo would have a strong financial incentive to make sure cars are patched so they'll make darned certain they are patched.
That doesn't mean we don't need audits and standards. It's just that we can rely on car manufacturers to create them, and to do a better job of creating them than would happen if they were imposed by legislation.
> someone in the forum pops up with "who's going to pay for it if/when they crash"
A slightly different take...
Who is going to jail if something goes wrong?
It's one thing to pony up for an insurance payout for property damage. A bunch of actuaries will do some impressive work in Excel and tell Volvo to set aside a few hundred quid for each car sold.
"Who is going to jail if something goes wrong?"
The same person who goes to jail when a plane crashes. If you can show negligence during design or manufacture, which is extremely unlikely, then the directors might have to carry the can because they are legally responsible for what their company does. In practice, you won't be able to show that and only financial penalties will apply, but that's probably sufficient to motivate a company because bankruptcy is death.
I will believe that when I see a director go to jail over cheating an emissions test. No, it will be the fault of some low level engineer and we had no idea and thank you for my golden handshake.
Let's be clear here. Company decision makers are swayed by stock movements not some bad event that their decisions today might cause in 5 years time. There will always be the pressure of penny pinching. There will always be some bug in the RTOS they build on, or some incidental scratched lens on a camera that confuses the software. I believe they will accept financial liability because the bean counters have figured that it is still a profitable venture. On a criminal level, they will deny it up hill and down dale.
"This means that it should be incredibly cheap for a private person to insure a driverless Volvo, and thus removes the biggest obstacle to takeup*"
It gets better. In the longer term, it will become increasingly expensive to get insurance to drive a car yourself.
Road accidents current kill a thousand or more people every year in the UK. The legal cost of accidentally killing someone is relatively low, because society accepts the argument that "everyone needs a car". Once it becomes clear that everyone can have a car at a much reduced risk of killing other people, those who insist on doing it manually will face stiffer penalties for their mistakes because courts will argue that they are negligent in not using the driverless option.
"At this early stage of development, it would be fairly easy to have small video cameras with outside views making a 5 minute (or whatever) looped recording onto an on-board SSD"
Recent Nissans already have the capability do do this, and if they made a model I liked I would have bought it.
Some people seem reliably to forget that an autonomous vehicle is going to be keeping a record of sensor input right up to a crash. If anything criminals are going to give them a wide berth; they will be mobile surveillance of a kind the police can currently only dream of.
Ah, yes. Any driverless car must have remote sensing, probably optical (audio, too?) and almost certainly that sensing data must be stored for some period of time, shorter or longer. Who is the owner of the data, then? Is it the owner/operator of the vehicle? Is it, e.g. Volvo "because we need the data for compliance"? Might it be The Government, who will make data access a provision of the licensing? Just who will be able to obtain a copy of the car's data, and in what circumstances? This is a question that I shall want to be answered long before I consider a driverless car of my own.
Dystopian vision: Google (tm) will license their driverless car technology to manufacturers, on condition that they get data on car journeys "to improve my travelling experience".
"Dystopian vision: Google (tm) will license their driverless car technology to manufacturers, on condition that they get data on car journeys "to improve my travelling experience"."
Utopian vision, Google is so large that it's worth it to give everyone free driverless cars because it'll significantly increase personal usage of google when people have an extra hour or two a day free.
Just who will be able to obtain a copy of the car's data, and in what circumstances?
AIUI, if a vehicle is involved in an accident or a crime, all the on-board ECUs/SatNav etc are fair game for plod to peruse as required. I'd expect that at this m oment in time they only delve that deeply for something fairly serious since plod are not renowned for the in depth IT skills and will most likley have to hire in expensive experts.
Over time I'd expect them to eventually to just plug into the car and have their systems build an "instant" VR simulation such that even your basic beat plod could operate it. That is when the real concerns over who gets access to the data becomes relevant. So it's good that we think about it and maybe set up legislation now so we don't end up with every plod involved in random pull-overs plugging into the data-port.
It's going to happen and there will be unintended consequences along the way.
I'm going to be the one driving around in a hacker-proof, modifiable vehicle from the early nineties until such time as non-autonomous vehicles are deemed to be incompatible with the robot world and are banned (hopefully well after I've shuffled off this mortal coil).
I wonder, are they going to incorporate the rules driverless cars follow into driving tests?
I'm guessing they won't be programmed to allow you to let one go against the road rules in a situation where you need to to be able to advance yourself based on the traffic behind it. e.g. mini roundabout driverless car and bus behind it next to a parked car, if you go you block the bus and yourself, if the driverless car goes the bus advances and you can now go, of course the bus should have waited behind the parked car or cars but in real life that doesn't always happen and if it's a row of parked cars going round a bend the bus has no way of knowing it's going to block people in. Also I wish them luck in predicting what a taxi/mini cab driver is going to do on the road, I still get surprised by them and I've been driving for 15 years.
It's a bit brave accepting liability, will that make car insurance a thing of the past or at least lower the premiums as you are now only covering for theft?
Easy enough to nick one but probably not worth the hassle (although those biiiig ass batteries are gonna be worth quite a bit, assuming they all end up being 100% electric)
If I wanted to take one I'd jack it up onto rolling dollys under the drive wheels and winch it into a fully metal bodied truck. Get out of that faraday cage, robot - mwahahaha... Car can't go anywhere and can't communicate with a server to alert the authorities to it's position (not that it would know that anyway - wouldn't be able to get the GPS - ho ho ho)
Like I say though, not worth the hassle (probably - how much are those batteries worth?)
"Car can't go anywhere and can't communicate with a server to alert the authorities to it's position"
It already sent the SOS long before you got it into the truck. It's amazing how quickly an SOS can be sent with GPS co-ordinates, even on a 2G signal. It's even quicker than Morse code. It might even have time to send full hi-res video of you jacking it onto the trollies if it's got 3g or even 4g.
...how does it tell the difference between a car stopped in a dumb ass position and a traffic jam?
How does it know the car in front, stopped in a dumb ass position in front has not stopped for a very good reason, such as a kid lying in the road after being hit.
How does it realise there is queuing traffic in lane 1 to get off the skip road, so you need to change lanes, instead of just sitting there politely for 45 minutes.
Then you have the issue, if you rely on the driver for only say 5% of the time, who's to say the driver will be paying any attention.
Hmmmm
"...how does it tell the difference between a car stopped in a dumb ass position and a traffic jam?"
A combination of identifying what is ahead and the information that it's receiving from other vehicles and traffic management systems.
"How does it know the car in front, stopped in a dumb ass position in front has not stopped for a very good reason, such as a kid lying in the road after being hit."
Irrelevant. It can still drive around the obstacles including anything in the road. (Only probably more safely as it won't be "rubber necking")
"How does it realise there is queuing traffic in lane 1 to get off the skip road, so you need to change lanes, instead of just sitting there politely for 45 minutes."
Again, this is down to working with the data available from management systems, other vehicles and from what can be detected ahead
"Then you have the issue, if you rely on the driver for only say 5% of the time, who's to say the driver will be paying any attention."
Various studies have shown that most drivers only concentrate on their driving for about 25% of the time. This is especially true on the motorway style roads. The autonomous cars will be focussed on the driving 100% of the time, will never get tired or bored. That has to be safer.
Yeah, this is where I see this falling over badly - Human drivers are bad at this as it is.
Case in point: On Sunday Oxford City Council, in its infinite wisdom, closed one of the major arterial roads into the city so that there could be a half marathon run along it. They diverted the traffic onto a small residential side street that runs parallel to this road. This street, also in their infinite wisdom, they filled with single-lane constriction points every 150 yards or so in order to "Slow traffic" a few years back. As it's residential, people are also permitted to park on it for most of its length.
Imagine what happens when a queue of traffic going in one direction hits a queue of traffic trying to go in the other direction on a road like that. All it takes is for one person to pull out into that little constriction without there being room for them to leave it again on the other side and the whole thing locks solid.
I was stuck in this for two hours, and had to get out of my car several times and go direct traffic - and I was by no means the only one doing that - because people kept trying to move into spaces that were gone by the time they got there.* The only way out was for chains of people - sometimes 10 cars long - to back up.
Now, in a perfect world with only driverless cars then this presumably would be preventable because they could all negotiate with one another and you wouldn't get that one dick head in an XC90 that decided everything was taking too long and repeatedly tried to force his stupid oversized piece of crap into places where anyone with eyesight could tell it wasn't going to fit and nearly taking people's mirrors off!**
The answer "That problem will go away once driverless cars are in the majority" or "Driverless cars will use telemetry from other cars" isn't good enough. It's like the bit where you open the box with the number 4 wrench (included in the box!) Until there is some critical mass of the things out there feeding one another data it just won't work.
Don't think I'm against the technology, but this is a hard problem, and there's a lot of handwaving when it's raised. Until someone has a very clear plan about how to get from where we are now to where we want to be, it's going to be hard to sell the things.
*And if the people organizing this wonderful public disruption had any brains they'd have seen this coming and sent some traffic police down there to do this professionally, but predictably, they didn't.
**Sorry, that's a rant, but that one guy was probably responsible for 45 wasted minutes on what should have been a 10 minute drive home
These problems are solvable though...the cars can work as a team and don't have to factor in impatience, deadlines or testosterone. In the case above, the cars can also fend off new traffic and reroute it; rather than (as currently happens) jams stacking up at both/all ends and making everything worse.
There's going to be many mistakes made along the way; but it'll happen eventually. I'm quite looking forward to a car that can take me on a pub crawl...dunno if it'll happen in my lifetime though.
You have my sympathy.
Here's an example of some similarly idiotic drivers. https://www.youtube.com/watch?v=_eKR-PCAJoI
I had a case some years ago near to where I live. There is a low railway bridge that prevents higher sided vehicles from taking the main route; this is clearly indicated along with the signs indicating the correct route for them to take. But even so, every month, some lorry driver will ignore the signs.
The one particular Saturday, it caused total gridlock around town and had been for about half an hour. This was because nobody would give way to anyone else. As it happens, I was walking in the area; I spoke to several drivers and after about 10 minutes made sufficient space to get the lorry backed up enough to clear the way for traffic to start moving again.
Within another 10 minutes, because the roads were clearing, the police could finally get there. I left them to organise turning him around to head in the right direction.
It wasn't difficult, but no-one seemed prepared to actually do a damned thing until I got there; then they were more than happy to be told to move over, back up a bit, make space etc. But no-one seemed to want to move until they were told to.
@JT
That's an interesting case scenario. As you say, these cars at least in first iteration need to operate in a world where most other cars are not self-driving or smart. Ideally they need to be working totally independently ie even without any outside communication to internet services or other external notifications from nearby cars / other objects, their onboard sensors and processing should be able to handle the driving requirements.
Can these cars recognise hand signals given by police officers directing traffic? Can they pass through a red light if it's a policeman who is waving them through? And, most importantly, can they tell the difference between a police officer directing them with hand signals and a random Joe doing the same?
In the scenario you describe, ideally a self-drive car would be able to recognise that someone is not just 'in the way' but also giving it signals to proceed, stop, back up, move this way or that. It should be able to do 'know' that in some cases this will involve breaking signalled traffic rules, and yet also be smart enough to not blindly follow any hand-given directions.
Recorded telemetry and sensor data in driverless cars (speed, GPS track, distances to cars in front/behind/oncoming, all-round video and radar/lidar data etc) prior to such crashes essentially rules out any profit to be had. That data is an insurance assessors wet dream and will prove categorically who was at fault.
That's if they can even get the driverless car to crash into them. These things will always react faster, out-brake, and remain at a safer distance from the car in front than any human driver.
Expect some attempts from the more stupid crash-for-cash tools, before they realise it'll be near impossible to crash into a driverless car and make it look like the driverless car is at fault. Then they'll give up and go on to easier pickings.
The crime problem for driverless cars isn't insurance fraudsters, it's when thugs learn to spot them and know with a large amount of confidence they can block the road and the car WILL stop (whereas a meatbag-controlled missile seeing a masked figure in the road might just put the foot down.)
I doubt it'd be a big problem, but I can guarantee someone will try it and the car manufacturers will have to consider it when building cars (possibly automatically locking the doors under certain conditions or somesuch.)
lol brilliant.. how often have you come across a masked figure in the road and had to put your foot down?
Most expensive cars these days have automatic door locks that lock when you drive off and you will be sat in a car, hands free so able to make a phonecall to the Police, complete with video footage of the masked figure lovingly archived by your friendly car.
Never; but that has no bearing because I am talking about a problem that could arise, not one that already exists. Perhaps a masked thug scenario is a bit too far fetched though; think instead of trying to get home through the town centre on a Saturday night when a bunch of drunks notice 'one of them new cars' and decide it'll be a laugh to see how long they can hold it up for?
"think instead of trying to get home through the town centre on a Saturday night when a bunch of drunks notice 'one of them new cars' and decide it'll be a laugh to see how long they can hold it up for?"
Yeah, I keep wondering about that. I was thinking a bunch of teenaged scrotes, but same difference. If you surround an autonomous car, what will it do? Can you herd it by leaving one side of the circle open?
The simplest solution might just be one-way windows, because what's the point screwing with an empty transpod?
Insurance companies would love you to have a safe driverless car.
They'll love paying out less.
Though it would be great to think this translates to lower premiums, corporate greed and thirst for ever greater profit means this is unlikely.
More likely premiums will stay the pretty much same for driverless cars, while skyrocketing for people who want to drive manually.
The only way premiums will drop is forcibly by legislation. With a corporate-friendly government in power, and large supply of cash-stuffed envelopes from the insurance lobbyists, don't expect to see that becoming law any time soon.
You're assuming that the current situation will prevail. If truly autonomous vehicles become a thing (and they likely will in time) then insurance for the driver is meaningless and all liability will likely be handled by the manufacturers, as Volvo are indicating here to an extent. The insurance co's will do big deals with manufacturers and it will all be rolled into the purchase cost.
You are also ignoring competition pushing prices down - insurance is a highly competitive sector. If company A has a policy for £1000 and company B can do it for £990 while still maintaining a margin then they will. Margins will increase even as prices drop due to increasing redundancy in the claims handling departments: logfiles from AVs will obviate the need to argue the toss about fault.
As an example of competition in vehicle insurance pricing having happened in my lifetime (and I think Arnaut the Less may confirm this as I gather he's a fellow British two wheeler) in the eighties there was only one company that insured motorcycles: Norwich Union, as soon as more companies started to offer motorcycle insurance, prices and cover options became more competitive.
The only time you'd need insurance for a (mostly) autonomous vehicle is when a meatbag takes over the controls and this eventuality will ultimately be ruled out as the tech improves.
Actually they make much of their money by investing the premiums you have to pay up front. They don't care whether it is a human or computer driver. The premium and excess will take care of any difference. They would be more worried about driverless fleets running Uber-esq services because it effectively becomes a monopsony.
A lot better, because these cars are actually paying attention all around *all* the time, leaving the passenger free to make phone calls, play with their laptops or even doze off without endangering anyone else which is a big change from the current situation.
If someone is run over and killed, their family is probably going to sue Volvo for a whole lot more than they would sue you or me. They'll claim "Volvo rushed their cars to marked, ignoring known safety issues" and try to tack on $100 million in punitive damages. Depending on the jury (in the US, other legal systems differ) they might even get it, since the safety record of self-driving cars will be quite short at first, and there's a long list of known safety problems buried by automakers - always something big in the news every year or two.
The record will become longer than a typical human will ever manage within months, maybe faster.
A typical human might drive for 70 years...
So if you sell 12*70 cars = 840 cars...
Then they'll take 4 weeks to accumulate a lifetime's experience - in fact the test program has probably already gained more than a lifetime's experience!
Added to that is the fact that the early adopters won't be the elderly ladies who run them down to the shops once a week, but the salespeople who do 35-50k miles/year, and want to make phone calls on the journey - the time taken to get HUGE experience will be trivial.
The problem isn't going to be the amount of experience, but the programming that applies it. Software is buggy, it could have the equivalent a trillion road hours of driving in every possible condition; it won't matter when the program applies that 'experience' incorrectly. Or sensor data is interpreted incorrectly, or wrong conclusions are reached about the condition of the road, or what have you.
Despite all the controls and testing of the software running airplanes experiences bugs, sometimes fatal ones. Look at what happened to that Mars lander that crashed due to metric/English conversion issues? That's what will trip up the self driving cars, and lawyers will know they have a deep pocketed defendant to go after. Look at all the people wanting to sue VW (in the US at least) for the lost value of their VW diesels.
Think about people who are against mandatory seat belt laws. They'll point to those rare accidents where someone is thrown clear and survives, but clearly would have died if they were belted in. Doesn't matter to them that 99% of the time being belted in is better. The same thing will happen with self driving cars, when they do have an accident that is due to a programmer error or sensor error it will be something where a human driver would have been fine. That's the Achilles heel.
Add in a coverup, which automakers love to do because they're eternally optimistic that the cost of making it public makes trying to cover it up worth it (makes one wonder how many SUCCESSFUL coverups there are if we keep hearing about stuff that was covered up and wondering why they'd do it when it harms them so much more than just coming forward when the issue is found)
I simply don't see "autonomous vehicles" happening, at least not as in fully autonomous anyway. IMHO it is too much of a leap to take in one go, which is what appears to be the target of the idea's proponents. It's as though the Wright brothers having managed to fly a few yards decided that the next step was to be Concorde.
I also cannot see how any serious degree of autonomy can be achieved without a great deal of data being passed to vehicles from roadside equipment; although my car and its SatNav are 5 years old and thus "old technology" as far as I can see the mapping data available to a driver is simply inadequate for autonomy to be possible. How will a car know what speed is permissible on any particular section of road? How will a car know that a reduced limit has been implemented for some reason?
One technical point has me wondering; what technology will be used by the vehicle to determine what is going on around it? At the moment I cannot see how data could be gathered without using radio signals, and if that assumption is correct the amount of spectrum that would be needed to ensure that different vehicles could not under any circumstances be confused by signals emanating from another vehicle. In addition, the spectrum requirements will be increased to give complete assurance that the RF transmitter on any given vehicle cannot "jam" adjacent vehicles simply by inadvertently overloading their receiver input(s). And how will the possibility of false returns be eliminated?
I am surprised by a couple of points; any El Reg thread about the IoT attracts a lot of adverse comments about the idea, and yet threads about autonomous cars bring out all sorts of comments in support of the suggestion*. Odd that, IMHO. The second point is that current technology would allow vehicles to be fitted with equipment that could help a driver (forward and rearward facing cameras to better determine culpability in the event of an accident being a case in point) and yet all the possible intermediate stages seem to be being overlooked in favour of the "single big leap" approach.
*I wonder how many of them are actually drivers? (I'll get my coat; it's the one with the Kevlar and ceramic plates in it...)
"I also cannot see how any serious degree of autonomy can be achieved without a great deal of data being passed to vehicles from roadside equipment; although my car and its SatNav are 5 years old and thus "old technology" as far as I can see the mapping data available to a driver is simply inadequate for autonomy to be possible."
As I understand it, having detailed-enough maps is currently the biggest hitch. We tend to think of pattern recognition as being harder because it's more technical, but in fact simply mapping all the roads well enough is a much bigger job because there are no shortcuts.
"How will a car know what speed is permissible on any particular section of road? How will a car know that a reduced limit has been implemented for some reason?"
The first one's basic mapping - current satnavs can do it. The latter, things like the Google car are quite capable of spotting and understanding signs like that. Even so, we might decide to modify signs by adding electronic beacons designed to be read by autonomous cars. At one time there were no paved roads, after all, and yet that didn't turn out to be a deal-killer for the (non-autonomous) car.
"what technology will be used by the vehicle to determine what is going on around it?"
Lidar. But let's run with the radio idea for a minute. Seems to me that you've specced requirements that are roughly the same as wifi. Hmm, the road network's going to become a network...
@DaveDaveDave
Thanks for your response; it's a pity that this topic has slipped down to the equivalent of the back page. I'm sure another related thread will appear soon.
In the meantime SatNav data can go out of date extremely quickly. Although I have never in 50 years or so failed to reach my intended destination (I can read a map!) I once checked the cost of updating my SatNav data, but at around £100 there was no way I was going to bother. It will have to be recognised that a new road, or housing estate or a change of speed limit is going to make any StaNav disc obsolete the day it is downloaded. Somehow data updates are going to have to be free and "instantaneous". How, exactly?
I was interested in your comment about Lidar; how well does it work in conditions of impaired visibility, such as fog, rain, or (perhaps worst of all) snow? (OK I know; humans don't do all that brilliantly sometimes) And how does any given Lidar installation work in the presence of a myriad of other Lidar systems around it, all completing for information in the same "volume"? Is any one installation guaranteed to be immune from the effects of any other?
I'm not sure about the comparability of my radio concern with wifi; in the grand scheme of things wifi is not mission critical, while vehicle safety most certainly is, or more accurately, will be. I suspect that any system is going to be vulnerable to receiver blocking (front end overload) or plain simple interference with messy consequences.
This discussion may have to be suspended until the topic makes a reappearance.
For what it's worth I don't have a dedicated satnav, I just use my phone on those rare occasions that I need direction assistance. I've a bog standard android and the Google maps seem up to date, even to the point of knowing about recent junction closures for works on Spaghetti Junction (I'm a brummie) and routing around them. Live traffic is also taken account of and alternative routes pop up whilst driving with the time implications highlighted to allow me to decide if it's worth taking an alternative route. The system is frequently updated to add new features too.
I suspect that when there are more connected/autonomous vehicles on the road they will be updating systems in real time and the more live data there is the more efficient they'll become.
Will we eventually get to a point where all vehicles are autonomous and network neutrality becomes an issue, with a fremium model allowing higher payers to be routed more efficiently at the expense of lesser mortals? Most likely with surge pricing too?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
