Reports: Autopilot will go on strike if you're not paying attention Rumours are emerging that Tesla is going to restrict the behaviour of its controversial Autopilot feature. The move follows criticism of the company after highly-publicised accidents, including the fatality in May when a car under Autopilot ran underneath a semi-trailer. Although nobody's offered evidence associating the …
<quote... aimed at the kind of people who think that their data is safe in the Cloud.</quote>
Are best utilized to impress upon manglement what can go wrong if you are not paying attention.
Picture this:
Your typical pathetic mangler, out driving around in his (or her) latest self driving car which has been hacked to accelerate at top speed into a brick wall, and the only way to stop the looming crash was to jam on the brakes. Mangler """trusts""" his (her) life to the self driving car, and gets a rude awakening.
"Picture this:
Your typical pathetic mangler, out driving around in his (or her) latest self driving car which has been hacked to accelerate at top speed into a brick wall, and the only way to stop the looming crash was to jam on the brakes. Mangler """trusts""" his (her) life to the self driving car, and gets a rude awakening."
I would think the results would prevent 'awakening'. With luck, anyhow. I mean, if I'm going to hack the gas pedal, I'm damnforsure going to not bother letting the car brake...
"music is muted"? My first thought is that it should be cranked to wake the driver up.
I'm also undecided on whether dumbing-down an interesting technology to suit the lowest level of driver stupidity is a laudable goal. Still, I guess it may save them from a speculative lawsuit or two.
I'm also undecided on whether dumbing-down an interesting technology to suit the lowest level of driver stupidity is a laudable goal.
It's not a simple case of driver stupidity, it's human nature - humans make lousy monitors of automated systems. Also the public roads are not the place to run uncontrolled experiments with "interesting" technologies, thank you very much.
Good follow up to the subject and makes sense, people can't be trusted.
European roads and some drivers are bad enough, It’ll be a long time before it works in Russia or a few other places around world where they seem to have a death wish or just bat shit crazy and think the accelerator pedal is an on off switch.
I wasn't sure if you were making a wanking joke or not, maybe because so many of your posts seem to get bogged down with politics. Just on case you didn't know, Fnaar is a Viz reference (whose appeal is mostly about puerile swearing and knob or fart jokes - hence the popularity on el reg)
The premise behind the Scientific American article is flawed¹ in that the environment (and I mean, in a wide sense, not just "plane in the air", "car on the ground") in which each technology operates is very different.
It is also a shame that people who I am sure are or were great professionals end up having their reputations tainted by being called an "expert" on the media and having nonsense soundbites put under their names.
Such as, quoted in the article:
"One obvious reason, they said, is that an airline or fighter pilot usually has enough clear air around them"
"Obvious"? There is very little we call obvious in engineering.
"Usually" --- Once, I told a first-time tandem jumper at the dropzone not to worry, because the parachute usually opens. He did make the jump eventually, after lots of reassurance, and I got a master bollocking from the DZ manager. But the point is, "usually" a car also has enough clearance around it, but you do not design something purely betting on the "usual"², which is why we "usually"³ carry a reserve.
If we are going to handwave "obvious" differences, to me one such is the thousands of hours required of training, study, and practice, before you get to get to sit on the right-hand seat of a commercial cockpit (i.e., 2i/c), plus the intense competition for it, as opposed to the very casual attitude towards car driving. Personally, that's were I would start looking for hypotheses to test.
¹ Maybe not as clickbait.
² You may choose to exclude extremely rare events from your design boundaries, but not the merely unusual. Especially not in safety of life systems.
³ BASE jumpers excluded, for practical reasons.
This post has been deleted by its author
You think the experience of very experienced engineers who've found out the hard way how difficult it is for highly trained humans to effectively monitor and supervise automatic systems is irrelevant to anyone trying to do something similar in a far more time-critical environment using minimally-trained individuals, because "road" not "air"? God help us.
> You think the experience of very experienced engineers
No. I think the quotes from Scientific American are contextless, irrelevant soundbites which, in being devoid of information, contribute nothing to their readership's understanding of the issues at hand.
I have no thoughts one way or another about the "experience of very experienced engineers".
I think you underestimate just how many fields of engineering NASA has experience in. The explanation from their experts has probably been dumbed down a lot (either by themselves due to the audience they had or by the reporter) but is largely correct. If in a flying incident, most of the time if you have only 1 second to react, you are dead. Inertia and aerodynamics mean there is just no way to change direction fast. On the road a second can be enough to change direction enough to avoid an incident. On top of that minimum separation in the air is usually several miles, several hundred meters at minimum for GA traffic. Land vehicles tend to operate within linguini to brontosaurus lengths from each other, making available reaction time that much shorter and collisions that much more likely.
The issue is, humans are not good at emergency situations when there is short reaction time.
Humans are at different abilities, some with a few hours of driving experience, some with many years. Many have not been faced with a major emergency situation in their life and it is not easy to practice as any simulation is controlled. Many are driving too close, do not react quick enough or over react to an incident. Even the best drivers can't be expected to know the exact surrounding and other vehicles every time there might be an incident.
A car can be made ensure that for 99.xxx% of the time it is always driving a safe distance (one where it can brake in time), it is always 100% alert and it chooses the best course of action in most cases based on multiple sensor and simultaneous inputs.
A self driving car can also get millions of hours of driving experience every hour delivered in regular updates by pooling the 'knowledge' of all other self driving vehicles and their situations. A human generally only gets their own knowledge by driving a car themself.
So the way it is viewed by those involved in the industry is that in nearly all situations a car can be made to be better at driving than a human (that is why there is some mandate for all new cars to have front sensors and emergency braking). There would be less accidents, less deaths, better traffic flows, less traffic jams. Even in aircraft often when a human takes over there are more incidents than when the computer is left to get on with it. They use autopilot because it is safer.
So the issue is when there is a failure of the car's sensors, it's control system or it faces a completely unknown situation. That is when humans expect to take over and when we feel we can do better than the car. Possibly we can, possibly we can't we don't know. This is why in the early stages we do need to be able to stay alert and not trust the self driving car too much, until we are confident in the technology.
However it is one of those problems that either, we as a society, need to be ready to say that we want to give it a go or we need to abandon the idea. Any half-hearted attempt will just run on for ever more until it runs out of money or the regulations get too onerous and the enthusiasm runs out. Seeing how long it took some manufacturers of cars to change over to fitting CD players as standard compared to tape decks I think it is good that others have decided to give this idea try.
A self driving car can also get millions of hours of driving experience every hour delivered in regular updates by pooling the 'knowledge' of all other self driving vehicles and their situations.
Is that really such a good idea? The result could easily be that a car faced with a given situation today may behave significantly differently to the way it behaved yesterday; the result will be driver confusion on a major scale. If the driver is always going to be legally responsible for anything that happens then the car must not be able to alter its operating parameters without the driver knowing about them and understanding them and knowing how to "interface" with them.
Then We haven't invested the time and money that Google, DARPA, BMW, Toyota, Daimler, Ford and a few others have but we don't think it will work... Well they might just be right. It would seem that NASA have done work on "human factors" and I would be very interested to know if the vehicle manufacturers have as well. I might even go as far as to say that NASA may have invested far more on that front than any of the car manufacturers; it would be all too easy to reduce driving to a series of 0/1 decisions and forget human factors completely.
"...a car faced with a given situation today may behave significantly differently to the way it behaved yesterday..."
Who said anything about radically different? Also why would you really care as long as it was better. If it gets feedback from human drivers that in nearly all cases a human will wait at a certain spot for another car to pass, the autodriving cars can also use that knowledge on that corner to adjust the way they drive. If they notice that on a certain section of road they are getting overidden to do something differently, it can be analysed and used to program the autodriver to do the same thing.
AC said: "A car can be made ensure that for 99.xxx% of the time it is always driving a safe distance (one where it can brake in time)"
But then some human bellend "jumps into" the safe distance space as lot's of drivers just treat "safe space" as a gap for them to use as they leapfrog down the road.
.. Spot the human who leaves a safe space and the space having a duration of sweet FA most of the time due to the above.
Well no, because sometimes an outside observer can see things which the insiders can't. For a good example, have a read of the Checklist Manifesto by Atul Gawande. It's about how surgeons trying to reduce deaths in surgery learnt from pilots about the importance of checklists.
This post has been deleted by its author
"One obvious reason, they said, is that an airline or fighter pilot usually has enough clear air around them"
"Obvious"? There is very little we call obvious in engineering.
The separation will usually be at least 3 miles laterally, and 1,000 feet vertically. The reason for the spacing is to give pilots sufficient time and space to react. It is obvious that pilots usually have sufficient clear air around them because that is what those regulations are for.
> The separation will usually be
With 2,500 hours of commercial flying I am (vaguely) aware of separation minima, except that those are not there "to give the pilots time to react", but to account for the measurement error of various bits of instrumentation internal and external to the aircraft.
This is a red herring anyway, since it revolves around the name coincidence between two systems which are vastly different past the fundamental concept of "it's a (bunch of) Kalman(s)".
On an irrelevant side-note, arguably, the car system is more befitting of the autopilot name, since it's the one progressing by means of visual references. The aircraft system should more properly be called "autohelm" or, when coupled to an FMS (Flight Management System), an "autonavigator".
"To be fair the lorry didn't have side protection for the trailer like we do over the pond in the UK so that contributed a bit too."
There is a good reason why in Europe the trailers have the side protection - it is because cars were running in to them and therefore they were being hit in an area with little protection and most likely to cause damage. So humans spend a lot of time running into the side of trailers anyway.
A simple solution, that would have saved a life would be to make all trailers in the US and other countries have side and rear protection so when a car hits it (whether computer controlled or human controlled) the occupants have a far greater chance of survival. But that is probably a bit too socialist for the USA.
"A simple solution, that would have saved a life would be to make all trailers in the US and other countries have side and rear protection "
This does show how it is not just about self driving cars.
The reason big US trucks don't have side protection is partly because the quality of roads in the more rural areas is so poor that there are lots of places where side protection would catch on humps. To make self driving cars work, roads must improve. Markings must be kept painted. Humps must be smoothed, potholes filled promptly.
The truck driver was making an illegal turn - better enforcement is needed. Recently a motorcyclist was killed near here because a Volvo driver decided to do a three point turn on a main road in a 60mph limit with the brow of a hill one way and a bend the other. Traffic law enforcement will need to be stricter to deal with the wetware-steered vehicles, as people will demand as close to zero fatalities as possible.
The whole system has to be rethought from top to bottom. But that's OK in the long run. A largely driverless car fleet will remove the incentive for willy wagglers, and reduce the need for car ownership, so there will be an economic adjustment away from making cars to making better roads for more comfortable vehicles, just as the move from expensive and inefficient steam trains to overhead wire electric trains reduced train operating cost while putting more investment into infrastructure.
Two car family here, one of which is only used occasionally (say 2 days a week). If I could send my car home again after I get to work, the first thing we would do is sell the other one.
We're rural, but if we were more urban, we'd probably be fine without owning a car, provided the cost of a shared system was significantly lower than the current cost of using taxis.
There are many, many people like us, who see a car as a useful but expensive tool, not a prized possession to be coveted, polished and drooled over. It gets me from A to B nicely, but costs a lot of money. If I can still get from A to B as conveniently, but for less money, I'll drop it in a heartbeat.
Recently a motorcyclist was killed near here because a Volvo driver decided to do a three point turn on a main road in a 60mph limit with the brow of a hill one way and a bend the other.
Stupid Volvo driver, maybe. But if it was a broken down Volvo driver we'd be saying that the motorcyclist had been riding too quickly for the distances they could actually see. There's never a guarantee that the bit of road you can't see is going to be as clear as the bit you can see.
"Stupid Volvo driver, maybe. But if it was a broken down Volvo driver..."
I don't disagree with your post - I would only observe that if the car had been broken down and occupying one lane the fatality would have been avoided. It was the being right across both lanes that seems to have been the problem. I don't know how fast the biker was going - we'll have to wait for the full report of the inquest - but the tendency of wetware not to allow for bends and humps is well known and it's just one of the areas where self driving cars would have an advantage. I know someone is working on cruise control that uses GPS to vary speed ahead of sharp bends and other hazards. I also speculate that a self driving car would not permit a U turn under potentially dangerous conditions, or at least would make very loud noises directed at the driver, while broadcasting its presence to other cars in the vicinity.
I look forward to the day when all wetware driven cars have a display on the roof which shows the number of licence points of the current driver. (Not really).
I look forward to the day when all wetware driven cars have a display on the roof which shows the number of licence points of the current driver.
Meaning nothing. I've no points on my licence and never have been in several decades of driving, and...well, there's nothing you'd be proud of in my history of reckless speed and careless misjudgements. The only thing I could offer the Court of Public Opinion (tm) in mitigation would be that my reckless speeding wasn't in OBVIOUSLY dangerous places, and that I've been out with IAM examiners who've told me that that on THAT occasion I'd have scored an easy pass.
(Not really).
Most certainly not. I wish I was a calmer, more prudent, more patient driver.
"Most certainly not. I wish I was a calmer, more prudent, more patient driver."
You get that way with both age and multiple experiences of being frustratedly stuck in huge traffic jams, slowly getting more and more angry until you reach the head of the queue and see the torn and mangled wreckage, the blue flashing lights, and maybe what is almost certainly a dead body being hauled into the ambulance.
Most traffic queues aren't like that, of course, but when you've seen a few over the years, it tends to make you think. YOU don't have to be the idiot to end up being the one in the mangled wreckage either. So keep a close eye on what all the other dipsticks are doing too. Sometimes it's not even a dipsticks fault. Saw a driver manage to control his car from a type blowout on the A1(M) the other week. Luckily he just managed to "gently" sideswipe a large lorry and damage a bit of the front nearside of his car rather than go under it. Considering the miles I drive, actually seeing an accident happen is pretty damned rare. I suspect he may have "damaged" his trousers too.
"I would only observe that if the car had been broken down and occupying one lane the fatality would have been avoided. It was the being right across both lanes that seems to have been the problem."
So replace broken down with "skidded out of control due to a blow out" or "and accident blocking two or more lanes". As a regular long distance driver I've come across both situations a number of times and you really do have to be prepared to stop within the visible distance. That can be a right PITA in Cornwall with single track roads, few passing places and 10' hedges either side which are nominally a 60MPH speed limit!
"The reason big US trucks don't have side protection is partly because the quality of roads in the more rural areas is so poor that there are lots of places where side protection would catch on humps."
Not really/ The cabs are longer and very low to the ground. Almost anything that would catch on a trailer with guards would also catch on the cab.
The issue seems to be opposition from the american haulage industry who feel that the cost to fit (maybe even retrofit) and redesign is too much for saving a hundred or so lives per year (how many casualties or life changing injuries on top?)
Even China has mandated the use of hi visibility side protection on trucks and their roads can be truly awful.
if you have to hang on to the steering wheel the whole time what's the point? you might as well go ahead and just drive the car.
Also "a driver who ignores the warning will have to stop their car and put the gear-stick into “Park” before they can re-activate Autopilot." Stopping the CAR in the middle of a major highway sounds safer. Although admittedly the driver is probably unconscious or dead seems legit.
So, did they flamingo-up the naming, or is it a sinister plan to improve the gene pool? The first thing they should do is to stop calling it 'autopilot'.
I don't think anyone expects auto-pilot / driving-assist / fully autonomous cars to be perfect; we will still have accidents for lots of reasons. However, the consensus is that we'll have a lot less accidents, and as we gradually perfect the systems the accident rate will continue to drop. Someone clever said 'science progresses one death at a time' and I expect it will be the same with this technology.
If the self drive system needs you to pay attention its not good enough, I imagine I'd stop paying attention very quickly once I'd got over the initial excitement. This is a backward step perhaps introduced as a placatory measure, but it can be no more permanent than having men with red flags walking in front of the car.
is to limit the maximum speed while autopilot is engaged. Under human control on the Interstate (motorway), I can legally drive 70 MPH (~113 KPH). Reality is usually 5-10 MPH above that, if one doesn't mind risking an expensive citation. If speed under autopilot were limited to, say, 50 MPH (80 KPH), that would allow more reaction time for the meatbag to resume control. Definitely NOT a perfect solution, but a workable one. And while I can't speak for anyone else, I would certainly be willing to occasionally trade some travel time for the ability to do something other than watch the countryside roll by. Not every time, but definitely some time.
That sounds reasonable. I'd add that the autopilot speed should default the legal limit of lorries/HGV/Trucks since they generally have the lowest legal speed limits on any specific type of road and usually drive at that limit (or are physically limited by a governor to 56mph with some adjustment allowed since most HGV speed governors seem to be set ever so slightly differently.
Does the US not have different limits for different vehicles? Are trucks and buses all doing 70mph along with the cars on freeways and interstates? Across Europe, trucks and buses have lower than the posted speed limits imposed on them. eg UK motorways posted at 70mph, coaches can do 60mph, trucks can do 60mph (but have 56mph speed limited fitted anyway) and neither of those are allowed in the outside lane when there's more than two lanes. On single carriage-way roads, trucks are limited to 40mph even on a signed 60mph area. Lower limits apply if your car is towing a trailer or caravan too. Likewise vans, so long as said van is not based on a car design. Similar laws apply across the EU.
Likewise, we have a "keep left unless over taking" rule (that would be keep right for most of Europe and the US of course) or does the US treat all lanes the same, available driving lanes? Obviously in busy/congested times, pretty much every lane is going slow, so all are up for grabs (expect for large vehicles as above)
"Does the US not have different limits for different vehicles?"
some places, yes, but not all
"Likewise, we have a "keep left unless over taking" rule (that would be keep right for most of Europe and the US of course)"
Again, some but not all places. As I have stated elsewhere in this forum, the US is not one homogenized unit with uniform laws that many outside the US are led to believe. The 50 state governments wield considerable power within their jurisdictions, including traffic laws. Some adhere to the common sense rules above and some don't. In a few remote places such as Montana and some parts of Texas and Arizona, the controlled-access freeways or toll roads adopt the Autobahn model - right lane for slower traffic, left lane has very high speed limit or none at all.
A relative of mine has an autopilot on his boat, very snazzy, set up some parameters (waypoints/speeds etc.) and away you go, has alarms on the radar/depthsounder etc.
Didn't stop him hitting a channel buoy and then running aground in the Humber estuary though, seems you actually have to put it into autopilot mode before it takes control.
(the boat had gone round several bends in the channel before the collision due to the tide/currents, it was steaming along with no one at the controls for nearly half an hour)
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
