Missing the obvious
The vehicles not driven by machines are driven by humans, if you apply the same rules that would not be allowed as many are not very good at it at all.
Experts have said the UK's guidelines for testing self-driving cars, published today by the Department for Transport, could put lives at risk. The new draft code (PDF) intends to bring trials to public roads in the UK. But unlike in California, the code does not introduce any mandatory requirements for the testing of …
"The vehicles not driven by machines are driven by humans, if you apply the same rules that would not be allowed as many are not very good at it at all."
And getting poor drivers off the roads is a problem because.....? (Yes, I know, people won't be able to get to work, their children will have to walk to school, they'll have to walk to the shops etc...)
Speaking as a very careful cyclist, I can't wait for autonomous vehicles *which obey the highway code* to become the norm. And no, I have no idea how you get from the current situation to an amazing utopia where autonomous vehicles, pedestrians, cyclists and manually driven cars can co-exist happily.
Yes there are lots of cyclists that ignore the highway code. There are also lots of motorists that consistently break the speed limit and jump red lights (crossing the line after the light has gone from amber to red). I'm a cyclist that (mostly) does not ignore the highway code; I certainly have lights, and don't cycle on pavements.
"crossing the line after the light has gone from amber to red"
The offence is crossing the line when the light is not green. There is a statutory defence if it unsafe to stop only while the light is amber. At least one cyclist does not know the law. (Also many other road users of all ilks.)
What he specifically said... and it is just above my comment, so I'm not sure why I'm bothering to repost it:
The offence is crossing the line when the light is not green. There is a statutory defence if it unsafe to stop only while the light is amber.
1. It is an *offence* to cross the line when the light is not green.
2. There is a "statutory defence" if it is unsafe to stop.
Goes on to deduce the commentor they are replying to does not know the highway code because they said this:
jump red lights (crossing the line after the light has gone from amber to red).
And therefore demonstrates how one cyclist is ignorant of the highway code, by implication all cyclists are ignorant of the highway code, and they can safely be run over, because it's probably their fault anyway.
Now try reading that lot again, noting:
1. It's not an offence to cross the line during an amber light if unsafe to stop.
2. Doing so is not a "statutory defence".
3. The original comment was clarifying the phrase "jump red lights" as about crossing the line after the light has changed from amber to red. Unlike the red herring about not understanding amber that has been introduced, this is unambiguous, and also an explanation of the phrase.
And on top, the highway code doesn't actually create offences. It can be presented as evidence in court. IANAL, and you shouldn't have to be one to drive, but my understanding is that offences are under the road traffic act, but details are set out by the highway code and various instruments, such as the The Traffic Signs Regulations and General Directions (2016) "An amber signal, when shown alone, conveys the same prohibition as red, except that, as respects any vehicle which is so close to the stop line that it cannot safely be stopped without proceeding beyond the stop line, it conveys the same indication as the green signal which was shown immediately before it." schedule 14, part 1, 4(9).
"I'm a cyclist that (mostly) does not ignore the highway code; I certainly have lights, and don't cycle on pavements."
And yet you want to punish all car drivers for the sins of a few, yet don't accept that many cyclist are far, far worse than you. Maybe cyclists should be banned from the road because of the idiotic ones?
Way to go with the sweeping generalisation!
I am a cyclist *and* a driver, they are not mutually exclusive, and I follow the highway code, rules of the road etc when in both roles. There are good and bad drivers and cyclists, so let's have less of that please.
How did we get from self-driving cars to slagging off cyclists?
"How did we get from self-driving cars to slagging off cyclists?"
This is a very common rhetoric in our British press comments section. Yes there are fucktards on bicycles. But somehow this then legitimises a possibly small subset, but very vocal amount of drivers to rant and threaten very real violence against all cyclists and state this is the reason why they can behave as they will. And thus the cycle continues.. :/
We are all equally allowed to use the road regardless of how quick our particular mode of transport will get us there. Therefore it is everyone's responsibility to ensure that other road users get to their destination safely.
Think about it.
If you are prone to epileptic fits then sadly you are likely not safe to drive. This is seriously inconvenient for some people and would actually be one great benefit of automatic vehicles.
The lights, yes, are there to tell people you're there. It's a legal requirement to have them after dusk (you may want to talk to those people asserting cyclists don't use them). Flashing ones last longer so you're less likely to find your batteries die during your journey and are easier to spot.
The fact that ultra bright LEDs and the even more ridiculous strobing of those lights that are available have reached the point that the are dangerous. This is not just for drivers, but pedestrians and other cyclists. These idiots then compound the problem by having them pointing horizontal so that from half a mile away you are already dazzled. Many of the lights have warnings about not looking directly at them however this advise only applies to the moron who is using it. I cycle, I have a good front light that is point downwards at the road (in the dark that is actually where it is needed). It is bright but it also has the option to dim it.
Add to that that many new cars are festooned with LEDs now, all far too bright (only headlights "should not dazzle oncoming traffic") this is becoming a real problem. The manufactures claim that it is for safety, that is bollocks, this is just one-upmanship in the endless specification wars.
Bright lights (headlights or DRL) are a menace, particularly as the average height with all the "SUV style vehicles" now puts them 3' off the ground to start with.
Sadly that utopia will never happen. AVs will never be able to cope with complex urban environments without the roadway being entirely segregated. The likely outcome, as with jaywalking laws in the US, will be for pedestrians and non-AV users to be corralled away from the AV roadway, in the same way that they're banned from motorways.
It seems to me that the best approach would be for a mandatory safety parameters to be baked in to all AV control systems but the issue there is that it would remove a differentiator for car makers (e.g. all cars would be the "Ultimate Driving Machine", not just BMW).
Speaking as a very careful cyclist, I can't wait for autonomous vehicles *which obey the highway code* to become the norm.
The irony is just too strong.
On my daily walk across zone 1 London, at best I see about half of cyclists stopping at red lights, often because the traffic has already began flowing across. At best I see about a quarter of them stop at pedestrian crossings when people are using them.
Sorry, but the average driver legally has to be trained to a higher standard than the typical cyclist. So you need to get your own house in order before bemoaning anyone elses.
Now, I will agree, the average standard of road use is appalling, but that spans the modes of transport, and while the typical car driver is abysmal, they are typically better than any other class of road user too.
So I gather that in the UK a lot of cyclists go through red lights. I assume that's after making sure there aren't any cars coming? Otherwise such cyclists would be an endangered species. While that's annoying and they shouldn't do that, how does that affect you as the driver of an automobile other than annoying you because you wish you could treat a red light as a stop sign like you see them do?
Cyclists have some pretty obvious disadvantages on the road since if they are in an accident that would be a minor fender bender to a car that will end up in the ER, and if they are in an accident that would result in injury to someone in a car they will be dead. So they have a lot more stake in road safety than automobile drivers do.
As an addendum... I not infrequently have to stop at green lights while cycling, because pedestrians start crossing as soon as the cars have passed. Heck, they start before the cars have gone and if I don't stop they'll get trapped between lanes. Some cyclists don't wait at pedestrian crossings, this upsets some drivers out of proportion to their views on, say, driving at 40 miles per hour down residential roads or texting at the wheel. This is not to say I have any sympathy for, to give an example, the twit a couple of weeks ago who cycled past a group of other cyclists stopped at a pedestrian crossing and who all of us passed in turn about thirty seconds later while he was having a chat with the police who'd also been waiting for the lights.
Perhaps you're unaware that in the UK cyclists are expected to obey the highway code when on the public roads. Part of that includes not riding through red lights A) For their own safety and B) for the safety of pedestrians who may be crossing. If you think riding through a red light is perfectly OK then you're part of the problem.
There's a difference between just running a red light without slowing, slowing down until you can see it is clear as if it was a yield sign, and stopping and then going through the red light as if it was a stop sign. I'm not advocating any of that (personally I always stop at red lights and wait for the green unless it has a sensor that my bike can't trigger, but I will blow through 4 way stops when I see no traffic coming)
Again though, I wonder why the anger toward cyclists? Are you jealous that they can run red lights and aren't getting nabbed by the cops? Were you hit or nearly hit by one as a pedestrian? Because you face absolutely zero risk of injury from a bicycle when you are in a car, so I don't see why you should care unless one is running a red light into the path of your car and forcing an emergency stop - and if so you should probably go easy on the poor soul as he won't have long to live doing something that stupid!
Were you hit or nearly hit by one as a pedestrian?
This. Literally every single weekday in London. Usually on pedestrian crossings, or when the cycle should be stopping at red.
The law is pathetically weak in this area. As a result, I now simply push them off whenever they get too close. They don't much like the road rash, which is hopefully disincentive enough to ensure they pay attention and stop cycling at pedestrians who have right of way.
They'd have to be within a few inches of hitting my knee with their pedal to make this viable; my arm only extends so far and I need to reach the cyclist before my arm is at full extension. Given the speed they're attempting to reach, that'll do too much damage to leave to chance.
I only fell the ones that come dangerously close - It's not like I'm out there knocking down every cyclist I see.
You may want to consider what happens if you badly injure someone by taking this course of action.
Nothing will happen - it's my right of way and they're hurtling towards me unable to take timely evasive action ensures that my self defence actions are justified and appropriate.
Had they given way where required, I'd not be able to push them off.
If they could take timely avoiding action, I'd not be able to push them off.
Were they a safe distance away, I'd not be able to push them off.
Sorry, but this will come down to self defence every time. If in the course of protecting myself the aggressor is accidentally inured, well, whose fault is that? The alternative of being crippled or killed by one of these clowns isn't appealing. Sorry.
Interesting thing about the highway code, there isn't a right of way, there's "give way to".
I'm not defending people riding through red lights, particularly if they stupid enough to do it while people are crossing, but if you're going out there with the attitude, "I'm in the right and I'm going to shove everyone out of the way," then you are a contributing factor to an accident. Are you able to take a timely measure to avoid them? If you're a road user then you should be thinking about avoiding accidents. Someone else is an idiot? Whether they're a cyclist or a bus driver you shouldn't be forcing an accident to punish them, it's not your job or your right.
Really? Because when TFL actually counted it turned out the figure was 16%. I try and avoid London full stop, but I reckon last time I drove there I saw a car or a scooter run just about every red light I saw, whereas I probably only saw cyclists run red lights on 10 individual occasions. So my made up statistic is motor vehicles 10 times more likely to run a light than a cyclist.
Speaking as a very careful cyclist, I can't wait for autonomous vehicles *which obey the highway code*
Speaking as a very careful car driver, I can't wait for cyclists "which obey the highway code"
See how that works? Not all car drivers are incompetent idiots. And not all cyclists are competent road users either. Tarring all with the same brush rarely works.
Would *I*, as a qualified driver, be allowed to operate a remote-controlled car on a public road, under normal driving conditions, entirely via remote control?
I believe the answer is no.
So why does an "AI" bot that's untested get to have that as the only safety measure if the electronics fail?
And, if I did do that, and it killed someone, would I be liable for not being in control of it - no matter if I couldn't override the controls or compensate even though I can see the accident coming - , or would the AI creator be held liable?
It's a stupid idea solved by simple testing procedures. Allow them with a supervisory driver inside the vehicle. When they prove themselves (but never needing any intervention), then have them do several thousand miles unaided (but with remote vehicle control). Then ramp up slowly.
But, to be honest, the REALLY stupid thing is the venue. When the devices prove themselves in a simulated environment (i.e. looks like a road, works like a road, not some on-screen fake 3D stuff) off public roads, and then on public roads but at slow speeds (e.g. moped speeds only), and then on motorways, and then in extreme conditions, etc. THEN you can authorise a full trial on public roads with no restrictions on what road/weather they can drive in. Not before.
I would be more than happy to let automated cars on our roads today. With a 20mph limit, not allowed on motorways, human driver behind the wheel. If they can't cope with that without interventions, they shouldn't be allowed anywhere NEAR 70mph, a motorway or the public, and certainly not unsupervised.
They're learner drivers, at best. Taking the instructor with dual controls out of the car before they've passed their test, or at very minimum a qualified driver happy to take the blame at all times, is illegal for a human at that stage, why should an untested technology leapfrog that requirement? Not to mention motorway driving, licenses with points on which they can lose, etc.
Assuming a test environment I have seen is indicative of the general case, the AI is training on conditions recorded from the road. This is video and depth information gained from sensors similar to those on a autonomous vehicle from an urban environment. It has poor road markings, shadowed areas from trees, etc.
There are weaknesses, for example all footage is in good weather, sun above the yardarm, driving on the right, traffic patterns are the same at each location; assuming the machine learning is as adaptive as AlphaZero, then different AIs will familiarize with local conditions so training for different legislative environments will be necessary.
I expect collecting large data sets from roads will be an early milestone for any serious automotive vehicle developer (or a supplier company will be created with a super-Streetview dataset). It would be interesting to see this data modeled and then a large number of AIs being run in it simultaneously in a MMORPG mode so they have to navigate around one another.
"It has poor road markings"
Considering the state of road markings on even some quite major roads here in the UK, especially since all the council cutbacks kicked in, I, as a very experienced driver doing about 45,000 miles per year, sometimes have difficulty working out where the lane marking are, especially in wet conditions at night. Computer "vision" is really not all that good yet so I'd be not at all surprised if an AV ends up in the wrong lane or even straddling lanes in those circumstances.
Just to expand on my last point, the MMORPG of AIs is to test ideas before they are implemented IRL. For example a novel road junction, a traffic-calming scheme, or larger ideas such as reducing the speed limit in a town, etc. and see how the AIs cope. Do they become calm and efficient drivers, or does some variant of Mad Max come to the fore?
There's a more fundamental case of missing the obvious. Why do the British government want to see experimentation on the British public?
Most of the UK motor industry is upping sticks, and was planning to do so before Brexit, due to high energy costs, high levels of bureaucracy, and high on-costs for labour. Any UK developed intellectual property will be privately owned and so sold, re-sold and rapidly be offshored to a low tax destination, or bought by a US tech corporation (or both). Even of the current UK car makers, they're all foreign owned, so the returns to the UK will be very slim pickings.
Sadly this is just more arts-graduates-politicians-picking-technical-winners. They've heard of AI, blockchain, autonomous vehicles, they must promote these things.
"There's a more fundamental case of missing the obvious. Why do the British government want to see experimentation on the British public?"
Off the top of my head:
- improved traffic management reducing congestion and allowing traffic to adjust to changing conditions (i.e. a blocked lane on the choosen route)
- reduced congestion reducing pollution
- potential safety improvements over self-drive (i.e. the removal of high risk activities such as dangerous overtaking, speeding)
- reduced costs/improved management of cargo logistics
- AV vehicles will likely increase the rate of change from petrol/diesel to electric vehicles benefiting the environment and helping meet pollution/CO2 targets.
- the security implications of AV allowing monitoring of potential suspects (equally scary from a big brother perspective, but you asked why does the government like it) and the potential to reduce the risk of vehicles being used in a dangerous fashion given the events in the UK/France/US/Germany.
None of these rely on owning the technology, but all present potential economic improvements over existing systems.
And I'm less pessimistic about the country benefiting from systems developed in the UK, even if they are sold. Getting some of a large amount of money is still better than getting none.
"- improved traffic management reducing congestion and allowing traffic to adjust to changing conditions (i.e. a blocked lane on the choosen route)"
This, and all your other "top of the head" suggestions are at least 20 years away in real life. No government minister is going to be think past the next election. They are acting like it's a done deal and just needs a bit of tweaking. Those of us who actually know something about computers, even if not in the AI or computer vision fields, know we are long way from real AVs
"This, and all your other "top of the head" suggestions are at least 20 years away in real life. No government minister is going to be think past the next election. They are acting like it's a done deal and just needs a bit of tweaking. Those of us who actually know something about computers, even if not in the AI or computer vision fields, know we are long way from real AVs"
You're pointing at the end goal and suggesting that it's 20 years away - you maybe right. However there are a lot of steps in between where more tasks are taken away from drivers, particularly in large cities.
This is stuff that a number of other countries (USA, Japan - I don't have details for Europe) are already doing to work towards better traffic management.
This is stuff that a number of other countries (USA, Japan - I don't have details for Europe) are already doing to work towards better traffic management.
In the most obvious use case of congested motorways or urban roads, AVs won't deliver much over what could be done with existing technology such as ramp metering, variable speed limits, phased traffic lights linked to volumetric sensors. And even then, the marginal benefits of traffic management are quickly exhausted.
Look at the UK's Managed Motorways programme. What it does, it does very well, reducing accidents, congestion and improving capacity. Unfortunately what any regular user of these motorways knows (and any idiot capable of operating Google Maps can see on a computer) is that the improvement buy a few short years - and sometimes not even that - before the demand outstrips that raised capacity. The M42 is a classic example, but it is also evident on the M25, M60, certain sections of the M4 and M5. Having AVs won't make much material difference to this any time soon, if ever..
Politicians and technologists are clutching at straws. Technology can improve utilisation around the margin, but it won't resolve the problem of transport systems whose physical capacity is saturated, whether that's road lane capacity, road interchange capacity, car parking capacity, bus system speed and capacity, underground and rapid transit capacity - including interchanges, payment and access control, escalator, platform and stair capacity. The best solution is not AVs or smart cities, it is simply relocating business districts closer to where people live rather than in city centres dedicated to commerce and business, and a dramatic increase in the use and effectiveness of remote working solutions.
"The best solution is not AVs or smart cities, it is simply relocating business districts closer to where people live rather than in city centres dedicated to commerce and business, and a dramatic increase in the use and effectiveness of remote working solutions."
To some extent, I agree. But my start-of-day travel, depending on where I'm going, is along or across one of the "corridors" between Newcastle, Sunderland and Durham. There is a large amount of traffic going in each direction because people live in or near one city and work in another. I'm not sure what can be done about that. It may help some if the business are on the outskirts, but that means re-jigging public transport. The vast majority of bus routes are designed around outer to inner. If you live in one part of outer and need to get to another part of outer, then you will often find the only option is one bus to inner and another to the other outer.
The best solution is not AVs or smart cities, it is simply relocating business districts closer to where people live rather than in city centres dedicated to commerce and business
This has long been the case in places like Detroit, where the urban center became unappealing. The problem is that instead of an inward-outward flow that can be served by public transit, you end up with people commuting from suburb to suburb, without enough people going in the same direction at the same time for public transit to be viable. And the traffic is still terrible because the road networks weren't usually designed with suburb-to-suburb commutes in mind.
Part of the problem is people don't necessarily pick where they live strictly based on commute distance. Usually they pick somewhere because the schools are good, or prices are low, or they like the neighborhood. They don't pick up and move every time they switch jobs or their employer decides to change offices. The old company town scenario where Bob's Widget Works locates in Outer Slowsville and then everyone who works for Bob builds a house in Outer Slowsville just doesn't happen anymore.
“In the most obvious use case of congested motorways or urban roads, AVs won't deliver much over what could be done with existing technology such as ramp metering, variable speed limits, phased traffic lights linked to volumetric sensors. And even then, the marginal benefits of traffic management are quickly exhausted.”
Looking purely at Europe.
The EU has committed to moving to electric vehicles by 2040, so there is going to be a significant change in how people view car ownership versus today. Ie. In cities, trends in the decline in car ownership and longer lease periods will continue while trends in people using car services for individual journeys or short term hires will continue. This means the majority of vehicles in urban areas will be replaced within 20 years and not necessarily with ownership of the new vehicles. There is also the impact on vehicle resales.
With more new electric vehicles coming onto the road, particularly as services rather than larger purchases, features such as GPS, collision sensors, collision avoidance and phone home will become common.
What can be done to reduce congestion? Target throughput/good put rather than speed. Provide sufficient intelligence in the vehicles to sync speed with traffic light phases, reduce minor collisions by the use of collision avoidance, restrict speeds and reduce the blocking of intersections to drive first wave. As this is used and improved, incorporate route planning to get to the point where journey times are more predictable. This will also help bus services as well as potentially helping people to choose other forms of public transport or commuting options other than cars. From here, I would expect centralised traffic management systems to begin to allocate you lanes/positions and monitor your journeys to ensure you comply with rules if you are self-drive. Semi-AV vehicles would operate based on sensors and control information from the centralised control points. There will be interchanges between “controlled” traffic and self-drive. Add in Uber-type services to improve utilisation of space by reducing the number of vehicles sitting in parking during business hours. For goods delivery, the route and parking can be pre-planned to reduce disruption while unloading vehicles or control what types of vehicles an go where at what times.
This type of model works for lower speed traffic and will likely reduce traffic speeds, which may help public transport initiatives. I believe the variance in motorway traffic makes it a hard point to start making these changes - until you have close to 100% AV vehicles in-place, high speeds mean mistakes can have significant impact, while the same changes in London where average speeds during peak hours in London is around 5-10mph making collision avoidance systems significantly more effective. For pedestrians, I would expect more controlled crossings and road furniture to discourage them from mixing with vehicles.
If this makes cars more expensive but gives you the option to be a passenger in a safer vehicle, the majority of people will become passengers and use their commute for other activities than driving. Move from car ownership to journey hire and you also reduce the number of mechanically unreliable cars from the road as you push maintenance to companies that can be regulated more easily than individuals.
Compare this model with what is available in Tesla (and others) today and what is being trialled by the Japanese manufacturers around central control systems.
Two points from your post -
1. Unless there is a safety and liability mandate attached, I would rather guarantee my safety by owning and driving my own car. I can see for-hire autonomous cars never being maintained beyond the minimum required to keep it running, not being even cleaned out, ect. If a car is delivered to you that is unacceptable for any of these reasons, and you reject it, will the company be required to send another, or will they just send the same one since it's closest? How many times can you reject before they stop responding, or blacklist your number? And finally, how long before they decide that the car must be run at max capacity, making it into a minibus while still charging you the single rate? This is already done by some taxi companies now, and it'll be a lot easier for a company to do this if they aren't face to face with an irate customer.
2. Autonomous or owned, the main problem continues to be congestion. Either people will be waiting on traffic or waiting on cars. Plus, cars will be empty half the morning and half the evening moving people to and from work. A solution to both would be requiring in-town businesses to adjust open and close times to spread their hours out. In other words, no more 8-5 for everyone. This would lessen congestion during rush hour across the board.
This raises 2 more points, will The Boss be understanding that you couldn't get a car early enough to be on time? And will you be required to pay just your time in the car or also for the car to travel from the city center to your house? How will you even know that the car even HAD to come from the city? The car might have dropped a mid shift worker off a block away, but you still get the charge.
A lot to work out still yet, and they haven't even gotten the damned things to work yet.
"...particularly in large cities."
From what I have read, large cities are by far the most difficult for AV to deal with (pedestrians, cyclists, varying road widths, poor road maintenance etc.) Motorways are easy in comparison and the drive to reduce freight costs will make this happen first.
"Motorways are easy in comparison and the drive to reduce freight costs will make this happen first."
You'd almost think a series of hubs or terminals on the outskirts of towns and cities, linked by dedicated transport routes for all the main freight, and then delivered into the town or city by small trucks and vans might be more efficient. We could call them "railways" and "freightliner terminals". The things on the "railways" could maybe even be driverless since they would be operating in a highly controlled environment inaccessible to the public or other unregulated vehicles.
This, and all your other "top of the head" suggestions are at least 20 years away in real life.
20 years is quite optimistic but would definitely be worthwhile aiming for.
It has been 10 years since HS2 project started and as far as I know not a sod has been cut for it.
"It has been 10 years since HS2 project started and as far as I know not a sod has been cut for it."
More than sod has been cut - significant works started in 2016 near me:
1. Show that all your air-side vehicles (buses, tractors, catering vans etc) can run autonomous. Nice safe pedestrian free controlled and mappeable environment without potholes (one hopes)
2. Build an autonomous supermarket trolley that can navigate Tescos on a Saturday morning without killing a higher percentage of shoppers than the normal attrition rate.
3. Then test it on the road.
4. But make the AI car pass a normal UK driving test before it is allowed out without an in car safety driver and L plates. The whole thing. Real examiner, real roads, 'stop when i slap the dashboard', 3 point turns etc.
It was never a "3" point turn, ever.
It had always* been "turn the car around in the road".
You lose by touching anything - the kerb, another vehicle, pedestrians etc - or by not making progress.
As long as you are getting somewhere each time you move the vehicle, you still pass the section.
That's never changed.
*Since it was first added to the test. There was a time when there was no requirement to drive backwards at all.
"Even the best put lives at risk when in control of a moving object substantially heavier than a human."
Well, that's a loaded statement if ever I saw one. Everything in life is a risk, it's a matter of quantifying and assessing the risk and deciding on the level that is acceptable. The majority of drivers are responsible and don't generally do stupid things. Most of the problems with traffic on the roads is the increase in numbers of vehicles and the reduction of traffic officers thanks to a reliance on cameras of various types (ANPR, Speed, Red Light etc.)
"Potholes are a real disaster – the AV can't spot them."
On some roads near me, even if you can spot them you cannot avoid them as roads are more pothole than OK surface and cannot steer a car to avoid them all (not that easy to navigate a level surface path on a cycle in some parts- so if 2 wheels and a short wheelbase a struggle no hope with 4!)
"Potholes are a real disaster – the AV can't spot them."
First requirement should be to place a substantial amount of money in escrow up-front from which to compensate the collateral damage on a no-quibble, no-delay basis. Keep that topped up or the trial stops.
Bypass the insurance companies altogether. The trialists can put their money where their mouths are. It's the least they can do. After all the rest of us don't get to choose our risks from this.
It's a fluffy bit of guidance written by some junior with an arts degree - high on promoting the popular idea but with no clue about any practicalities.
It's an actual active hindrance to doing anything sensible. And just wait until there's a backlash when someone doing a test 'in full compliance with DfT requirements' wipes out some poor innocent(s) by remote.
The big car industry players will probably be sensible regardless, but some of the others might do something really really stupid in their fog of blind faith and 'disruption'.
Or possibly it will wait forever for you. It depends on how the AI interprets the scene.
One test program in the US had problems with their cars being blocked by seagulls. The seagulls weren't intimidated by the cars once they realized they wouldn't get run over, so they'd hang out in the middle of the road until the safety driver took over and inched forward.
The guy on Radio4 at lunchtime was saying that 6 cars would be tested in Greenwich later this year. That makes it sound like they already have the cars in mind to test.
But why do the first tests on small city streets? Surely you start out on something big and predictable like the M1. Set it going at 56mph and it's a (relatively) simple task. Certainly compared to parked cars, kids, cyclists, pigeons, and that bloody great line they set the clocks by.
Regardless of the ridiculous politics and bungs and back-slapping, I simply cannot see AV working any time soon - at least not in a real sense.
The whole notion seems to rely on "AI" ignoring the fact that despite all the recent noise about it, is still many years off being "intelligent" and in 99% of cases is not 'I' at all; it's just "normal" software. The basic problem of navigating around the streets, avoiding an infinite list of random hazards and sensibly handling an infinite number of situations is difficult enough for a human driver. There is no way the current art of computer software can do likewise; just getting computer vision to make sense of stuff is a herculean task in itself. And that's before you try and do anything with the information. Throw in some snow that obliterates what you're looking at, some random guy that decides it's a good idea to do a dodgy U-turn at a moment's notice and a crowd of people racing across a zebra crossing and wandering in 6 different directions and the problem starts approaching unsolvable.
Given the computing power of something like Google or Amazon, you can probably brute-force your way through some of the problems. But that kind of power won't fit in the dashboard of a car, and won't solve all the issues even if it did.
I'm not saying it will NEVER be solved. But I may not be around when it is
Its already been demonstrated that "safety drivers" will tend to be minimum wage peons with more eyes on the phone than the road, at least once the thing gets vaguely competent. Not that I'm presenting a solution, but there's a definite trap there that needs to be avoided somehow.
Reading the article it would appear a responsible engineer has been nowhere near that document. It appears to have been written by a politician or HR specialist. They should be kept as far as possible away from this stuff until it's ready. Engineering trials are serious business and media specialists have no experience of life and death decisions. Quite shocking, but then if you have a politico in charge, with no life experience, what can you expect?!
"The entire purpose of a safety driver is as a backstop for catastrophic technology failure," said Tindell.
Quite. And if the technology failure is that the system has crashed or seized up so badly that the remote link no longer works, just exactly what is the remote safety driver supposed to do?
Or the incredibly rare circumstance that the car is under a bridge, among tall buildings, or in a tunnel so that the link doesn't work?
"just exactly what is the remote safety driver supposed to do?"
For that matter, as most gamers already know, watching on a screen, reading the "instruments", is no substitute for the "feel" of actually being there. Taking a corner and the road surface is a bit frosty or just wet and greasy, and the "safety" driver has no awareness that the back end is starting to slide until too late. S/he can only rely on the so-called AI to cope in that situation.
In reference to the recent incident where you were correctly parked in a designated parking area:
The other party has reported that there was no way their autonomous vehicle could have known your car would be present. Therefore you are being held entirely responsible for the collision and will have to pay all costs accrued. Furthermore in view of your obvious recklessness, we are no longer prepared to provide you with insurance cover - with immediate effect.
Now come on, who here wouldn't love to draw a curve onto the divider line up to a brick wall and paint a tunnel on it as a test of this technology?
Yeah sure but you'll feel a right cunt when the AV goes straight through the tunnel and when you try to follow it you smack right into a wall
(Am a software engineer.)
Got a 2018 Nissan Leaf with ProPilot, not much different than Teslas features really. It can follow other cars, and does so very very well. It also avoids collisions, first alerting the driver with beeps and vibrations and then using the brakes automatically. However, the self steering is very unreliable to the point of being dangerous. For example, the vehicle will be driving along a motorway, and it will then disengage, causing the car to swerve towards either the lane to the side, or central reservation. Worse, it will disengage without reason when going around a curve, so the vehicle will literally, keep going straight. Of course, in all cases, I have my hands on the wheel, but then, what's the point?
There are multiple social, technical and geospatial reasons for the issues with AV. Would require a white paper to discuss!
So I appreciate you experimenting with other people lives - a controlled test? Are you really in control - hands on the steering wheel not withstanding? We have a Leaf - I would never trust any of the automated driving facilities - and they have a lot more money to throw at the problem than Tesla.
But it is a nice car - treated as a car with an electric drive. Bugger automation.
Having read the cyclist comments, I'm not convinced that a few deaths is a bad outcome for them.
It seems they broadly expect to die and yet carry on, so presumably they're happy with that.
As for the remoteness of the human fallback, where else would they be?
You aren't seriously going to prefer the guy in the back of the "autonomous" car are you?
Cars should be more efficient than a human at selecting the correct gear. The tech has been around for decades. These days it could scan the road and be better than just relying on current speed. Cars used to have a choke - now they don't, we got over that. We could even lose the accelerator as well as the gear pedal. Just have steering and brake, if you don't mind the car choosing your speed for you based on the conditions and the speedlimit.
I can't see true AVs ever working for the general case unless every car is an AV. Maybe for certain restrictied applications like a taxi on private land - or a Mars rover. There will always be people who want to drive themselves, which will confuse the AVs. But maybe self-drive won't be an option one day.
What I can see is more cars with driver assistance tech to make driving easier for the human (automatic parking, emergency breaking, lane following, heads-up displays - all available today). Some good tech may come out of AVs, but a car that can drive itself on roads anywhere on the planet? Probably not.
Where are all the autonomous trains? (OK, there's the docklands light railway) It's a much easier problem to solve. Get that done first, then fully autonomous planes, and then we can think about cars.
It's inconceivable that governments worldwide would allow the dangerous practice of allowing prototype AVs to be operated on public roadways when no minimum safety, security, engineering or maintenance standards have been established for these vehicles. Insurance companies are even refusing to provide coverage because the judicial and legislative communities have not established who will die or be injured in an unavoidable accident and who will pay when said accidents occur.
I think it only fair that 'remote' drivers should feel real pain if the car in their car crashes. It shouldn't be too hard to rig up an airbag with a few shards of glass and a couple of steel bars that gets triggered if the car crashes. Or really - as they are just employees - maybe the company bosses should feel the pain, we could put the 'tru-feel T.M' airbags in the coasters for their Gin and Tonics.
I love the comment that humans have to pass a driving test and the thought of a driving examiner sitting in the passenger seat and telling the vehicle what to do.
If it was me, I would want the Minister of Transport sitting in the back seat as well.
Autonomous vehicles are a ridiculous and total waste of money.
There might be a case for a very limited use for lorries on motorways and to be able to get to a depot close to a turn off but that is all. Okay, maybe for a coach in similar circumstances but I would still prefer a driver.
I do wonder how I might react when waiting at a roundabout when an autonomous vehicle comes round. Do I sit and wait or now that it will stop if I jump out?
Perhaps, somebody can answer a question. I know that bikes are meant to be given a 1.5 metre clearance. Does this also apply to horses and pedestrians and should bike riders be held to the same rules?
I might be wrong but I also have a feeling that pedestrians have the right of way whatever the lights say, certainly if the autonomous vehicle if it is programmed correctly would stop much to the annoyance of other road users behind.
Not much has been said so far of a remote control operator who can intervene in an emergency.
Is there any legislation, technology, training, test, insurance etc. for a driver who uses remote control?
Should they not have to demonstrate that they can drive safely in all circumstances via remote control before they are given the responsibility of overseeing an AV and intervening swiftly safely and appropriately should the AV encounter a problem?
Really, ffs why?
There are enough problems having a meat sack in the car as a safety override without trying to replicate the immersive experience of sight, sound and feeling that the driver in a car gets, using remote control technology.
You still need one on one mapping between supervisor and the AV and this is just adding another layer of technology and expense.
Unless, of course, some bright spark thinks you can have one remote human to every 20 cars, and the car will be trusted to ask for help then wait for the call centre to have an operative come free.
Highly trained local drivers from Mumbai will be able to add value and simultaneously drive costs down, of course.
The DVLA used to be a responsible organisation, until they started to use software to control road traffic on so called Smart Motorways. Their attempts to do so leave me with little if any confidence in their proposed self driving cars rules. How often do you drive along a motorway at night, where roadworks are in progress, and find the gantry signs directing one to use the nearside lane, while the road signs say use the outside lane. Complaints to the DOT are always refuted. About par for the course.