Thatcham research lab Tesla test - vehicle ahead takes avoiding action

[wraper]

So what? How does it help more deaths courtesy of the autopilot? Do you realize these people wouldn't be dead now if it were not for the autopilot?

That you way overblown the problem. It may be an order of magnitude safer than car driven by human but you are sill outraged as sometimes shit still happens.

Do you realize these people wouldn't be dead now if it were not for the autopilot?

They might be dead because of their reckless driving even without autopilot. Actually overall more people likely would be killed.

[GeorgeOfTheJungle]

Your logic is flawed.

[wraper]

Your logic is flawed.

Then let's ban all cars and planes altogether. They cause deaths!!!
Personally I'm fine as long as technology progress makes life safer than it was. I don't agree with a stance that it's acceptable as long as human deaths are caused by humans and completely wrong when rare death is cased by imperfection of automatic machine (humans still were responsible of negligent usage of technology).

[julianhigginson]

So what? How does it help more deaths courtesy of the autopilot? Do you realize these people wouldn't be dead now if it were not for the autopilot?

There's a stat going round that tells us Autopilot-equipped Tesla cars travel around 320 million miles per fatality, on average in the US, all cars (including teslas) get 86 million miles of travel per fatality. So, yes.. a small number of people have died due to autopilot, but statistically, a lot more have lived. We just don't know of them because they're not road statistics, like they would have been in another car!

YES, autopilot has killed people that would be alive today if not for autopilot (that said, the list of autopilot deaths includes that guy who was WATCHING A MOVIE while his car drove him to work, it got confused by the lack of contrast between the side of a truck trailer and the sky, and tried to drive under the trailer..... after that is when they started using the radar a lot more... I'd say that guy would have died a lot sooner if he'd tried that in a regular car!)

BUT autopilot is also very good in a lot of situations, avoiding very bad crashes because it has response times far shorter than humans, and has rules that cover how to best react in situations we may never see.

Automating driving is a real world example of the trolley problem, but executed statistically rather than directly. You can make a choice to save a large number of people overall, but that choice is going to kill a smaller number of people who wouldn't have been killed in a car crash otherwise.

[Marco]

If they don't fix its behaviour with cars in front of it evading obstacles and a human is one of the obstacles next time, Tesla is going to be as fucked as Uber was ... maybe moreso given how much their brand is tied to the convenience of using an electronic device while "driving". There will be no room for excuses, the behaviour is known, the lack of attention of the drivers caused by lane following is known, it's just a question if the technology improves in time for them not to get fucked by chance.

You'd best not be long Tesla if they end up on the wrong side of the statistics.

PS. about radar being better with rain, that's true. The problem is how to get high directivity with a small system. You can focus light with an optical system a couple cm on each side. With radar, not so much ... as it stands the radar doesn't really image anything. It just picks out doppler signatures for a huge FOV and makes assumptions. Potentially lethally wrong assumptions. I personally believe stereoscopic should be able to be a decent alternative to LIDAR, but regardless of whether you use LIDAR or stereoscopics, interpreting the images beyond just following lines on the road is no small task. Not a task which should be experimented with with inattentive drivers IMO.

[Howardlong]

Before we fall into the trap of bringing in aviation into the mix as a comparison, please remember that almost all commercial passenger transport is in controlled airspace: that means that traffic is proactively and centrally coordinated. Airpsace also has the distinct avantage of having a third dimension, there are no physical road constraints, and there's far less traffic density in those three dimensions. There aren't too many pedestrians either.

This is completely different to the road vehicle scenario where there is no proactive traffic coordination, centralised or otherwise, vehicles are constrained by roads into a far higher density, and any deconfliction is purely reactive.

[Mr. Scram]

There are terrible drivers and there are excellent drivers and everything in betweeen. In any case I prefer to be killed by a human error than by the fake "autopilot" pile of crap of a narcissistic megalomaniac named Musk or whatever.

There's never going to be, in a very, very looong time, any "autopilot" that comes even close to a good human driver because the technology and know how to pull that out does not exist yet. At least not as the streets/roads are now, not without help from additional support infrastructure, in which case it won't be an "autopilot".

The best you can say of the autopilot is that it has killed dozens already. Funny thing is in Elon's words it's never, not even once, been the autopilot's fault. Eat this. And if it has not killed more people it's because there was a human driver overseeing it.

You obviously have very strong opinions in regards to Tesla and advancements in road technology. Do you have any numbers to substantiate your claims, like on how autopilot systems are impossible for "a very, very loong time" for example?

Again, this is a statistical matter. Obviously people are going to be freaked out by relinquishing control to a computer, but emotions aren't or shouldn't be the deciding factor. Considering it's not just your own safety but those of others too, the decision is likely to be taken out of your hands. Spilling blood over silly emotions is ridiculous and criminal.

[Mr. Scram]

Your logic is flawed.

His argument is sound. Not liking an argument because it doesn't line up with your emotions doesn't make it flawed. Ultimately the statistics determine what the way to go is and those are in this early stage already in favour of automation.

[GeorgeOfTheJungle]

I think you don't understand the statistics.

[GeorgeOfTheJungle]

You obviously have very strong opinions in regards to Tesla and advancements in road technology.

I wouldn't call a device that's killed dozens already an "advancement in road technology".

[Mr. Scram]

Before we fall into the trap of bringing in aviation into the mix as a comparison, please remember that almost all commercial passenger transport is in controlled airspace: that means that traffic is proactively and centrally coordinated. Airpsace also has the distinct avantage of having a third dimension, there are no physical road constraints, and there's far less traffic density in those three dimensions. There aren't too many pedestrians either.

This is completely different to the road vehicle scenario where there is no proactive traffic coordination, centralised or otherwise, vehicles are constrained by roads into a far higher density, and any deconfliction is purely reactive.

I disagree. Commercial aviation is a mix of complex three dimensional no fly zones, air corridors and open areas. Self regulation also is a vital part of the mix that makes things so safe. The skies are much more constrained and complex than you'd think, while the situation on roads is in many cases more clear cut by design.

[Mr. Scram]

I think you don't understand the statistics.

This is the second objection you make in this thread without explaining how or why. I hope you understand that people here have no other option than disregarding opinionated statements without arguments or substantiation. It's no use guessing why you might object. If you formulatenan argument, it might further the discussion.

[Mr. Scram]

I would not call a device that's killed dozens an "advancement in road technology".

Most people would agree that less deaths is a better results. I gather that people who feel the world is overpopulated might disagree, but I don't expect many people would feel this is an appropriate point of view.

[Howardlong]

Before we fall into the trap of bringing in aviation into the mix as a comparison, please remember that almost all commercial passenger transport is in controlled airspace: that means that traffic is proactively and centrally coordinated. Airpsace also has the distinct avantage of having a third dimension, there are no physical road constraints, and there's far less traffic density in those three dimensions. There aren't too many pedestrians either.

This is completely different to the road vehicle scenario where there is no proactive traffic coordination, centralised or otherwise, vehicles are constrained by roads into a far higher density, and any deconfliction is purely reactive.

I disagree. Commercial aviation is a mix of complex three dimensional no fly zones, air corridors and open areas. Self regulation also is a vital part of the mix that makes things so safe. The skies are much more constrained and complex than you'd think, while the situation on roads is in many cases more clear cut by design.

What is it you’re disagreeing about? As I stated, almost all commercial passenger transport is in controlled airspace. It’s like that for a reason: safety by centralised coordination. There are many aspects to flight safety, and this is how almost all commerical passenger flights avoid bumping into each other which is the scope of what we’re discussing here, compared to the lack of centralised coordination an autonomous car is subjected to.

(And FWIW, I’m a pilot, based within the London CTR, I am sure you’ll know just how complex that airspace is).

[wraper]

Howardlong you missed the point of me bringing up planes. It's was mentioned to show that this technology causes deaths as well. There is autopilot too (since a long time ago) but it wasn't the point and was not mentioned. It was mentioned to show that it's hypocritical to put one particular technology (which increases safety) under tough scrutiny while there are other technologies which cause deaths and everyone is OK with that.

[Howardlong]

Indeed, and my point is that we should be wary of comparing commercial aviation (which has enjoyed centralised coordination for very many decades) to driving an autonomous car (lacking any centralised coordination) has very fundamental flaws.

Now if there was an effort to integrate some kind of centralised vehicle coordination longer term, that would be interesting, but it would also need to have legacy integration.

Tesla calling it an "autopilot" hardly helps!

[GeorgeOfTheJungle]

This is the second objection you make in this thread without explaining how or why.

This:

There's a stat going round that tells us Autopilot-equipped Tesla cars travel around 320 million miles per fatality, on average in the US, all cars (including teslas) get 86 million miles of travel per fatality. So, yes.. a small number of people have died due to autopilot, but statistically, a lot more have lived. We just don't know of them because they're not road statistics, like they would have been in another car!

Is a (textbook example of a) non-sequitur.

[wraper]

This is the second objection you make in this thread without explaining how or why.

This:

There's a stat going round that tells us Autopilot-equipped Tesla cars travel around 320 million miles per fatality, on average in the US, all cars (including teslas) get 86 million miles of travel per fatality. So, yes.. a small number of people have died due to autopilot, but statistically, a lot more have lived. We just don't know of them because they're not road statistics, like they would have been in another car!

Is a (textbook example of a) non-sequitur.

Is a (textbook example of a) non-sequitur.

[Mr. Scram]

What is it you’re disagreeing about? As I stated, almost all commercial passenger transport is in controlled airspace. It’s like that for a reason: safety by centralised coordination. There are many aspects to flight safety, and this is how almost all commerical passenger flights avoid bumping into each other which is the scope of what we’re discussing here, compared to the lack of centralised coordination an autonomous car is subjected to.

(And FWIW, I’m a pilot, based within the London CTR, I am sure you’ll know just how complex that airspace is).

The point is that the road also has all sorts of mechanisms to ensure separation and flow control. Just the concept of lanes keeps cars laterally separated. Intersections are either vertically separated or centrally controlled by traffic lights. Maximum speeds ensure relative speeds are limited. Stopping in high speed areas is not allowed. The flow of traffic on a more macroscopic scale is often controlled in busy areas by the authorities which can divert traffic, add lanes or close roads. Instead of the tower telling you to cross the runway, there's a traffic light. There really isn't much of a difference when you boil things down.

Conversely aircraft have collision avoidance mechanisms in place too, either in the form of electronics or obviously a pilot.

[Mr. Scram]

This:

There's a stat going round that tells us Autopilot-equipped Tesla cars travel around 320 million miles per fatality, on average in the US, all cars (including teslas) get 86 million miles of travel per fatality. So, yes.. a small number of people have died due to autopilot, but statistically, a lot more have lived. We just don't know of them because they're not road statistics, like they would have been in another car!

Is a (textbook example of a) non-sequitur.

Once more it would help to explain why you think that is the case, because it seems to be a sound argument. We shouldn't be left to guess why you disagree with the quote. Without knowing what you are disputing it's hard to have a well argued discussion.

Although I suspect it's more of an emotional matter than rational. The former obviously defies explaining. Somehow people tend to prefer to die at the hands of a human, even if that ultimately hurts their chances of survival.

[julianhigginson]

Once more it would help to explain why you think that is the case, because it seems to be a sound argument. We shouldn't be left to guess why you disagree with the quote. Without knowing what you are disputing it's hard to have a well argued discussion.

Although I suspect it's more of an emotional matter than rational. The former obviously defies explaining. Somehow people tend to prefer to die at the hands of a human, even if that ultimately hurts their chances of survival.

George of the jungle is obviously very triggered by the idea of giving up his god given right to kill himself, his family and strangers with his own natural human incompetence in relation to long periods of solid concentration.

He's resorting to textbook discussion derailing techniques, by pushing out the bare minimum of information beyond "I disagree" - this is because the less he claims, the less he has to actually back up, and maybe that's a good thing, seeing he's already failed to back up any of the claims he's already made.

[Howardlong]

The point is that the road also has all sorts of mechanisms to ensure separation and flow control. Just the concept of lanes keeps cars laterally separated. Intersections are either vertically separated or centrally controlled by traffic lights. Maximum speeds ensure relative speeds are limited. Stopping in high speed areas is not allowed. The flow of traffic on a more macroscopic scale is often controlled in busy areas by the authorities which can divert traffic, add lanes or close roads. Instead of the tower telling you to cross the runway, there's a traffic light. There really isn't much of a difference when you boil things down.

Conversely aircraft have collision avoidance mechanisms in place too, either in the form of electronics or obviously a pilot.

The primary method of separation for commercial aircraft movement is via proactive centralised coordination. As I stated already, deconfliction for road vehicles doesn’t have this kind of coordination, instrad it’s piecemeal reactive. There is no communication between all road users to allow for proactive centralised coordination, we jump in our cars and go. If it did have some means of universal proactive coordination between road vehicles, we wouldn’t be having this conversation.

Aviation collision avoidance systems, which do bear similarity to vehicle assisted automation systems, are reactive and very much a secondary method of separation. They are also relatively rare to flag, because the primary proactive central coordination method has already done its job.

Edit: I’ll just add ghat, as an aside, if there was universal centralised coordination model for road vehicles, road utilisation could be far higher and delays far lower. Practically speaking, it will be a long time coming, and not all road users will adopt it anyway.

[Mr. Scram]

The primary method of separation for commercial aircraft movement is via proactive centralised coordination. As I stated already, deconfliction for road vehicles doesn’t have this kind of coordination, instrad it’s piecemeal reactive. There is no communication between all road users to allow for proactive centralised coordination, we jump in our cars and go. If it did have some means of universal proactive coordination between road vehicles, we wouldn’t be having this conversation.

Again, I object to the notion of it being piecemeal reactive coordination. The system is much more coherent and refined than that. It's mostly the implementation that differs, but road traffic is centrally coordinated. The fact that cars travel in the same direction in very close proximity to each other at very similar speeds on a highway is no accident, it's by design. One might call that passive centralised coordination, but traffic lights and other more macroscopic flow control aren't even passive.

[Howardlong]

Any road traffic coordination is by nature reactive, vehicles appear at random times in an unplanned manner.

Now, yes, you can sequence trafffic lights to improve road usage in a localised manner. This is hardly the same degree of control that an air traffic service mandates inside CAS. For example, you can’t randomly choose to go off piste and turn left or right without first coordinating that with the controller, and such requests are not that uncommon, particularly in bad weather. Sure, a mandatory TCAS resolution advisory will demand immediate action, but that is very rare, and you’d want to be communicating any action with the controller ASAP. After all, ATC in CAS is already there to maintain separatiion. A TCAS RA inside CAS will generally be as a result of pilot or controller error, and that’s the reason it’s rare.

So what I am trying to convey is that autonomous vehicles are relying on a reactive system for collision avoidance.  You can talk about roads, lanes and traffic lights, but that is nowhere near the level of control that occurs in CAS. Outside CAS, yes, it’s not a million miles off, where we adhere to the Rules of the Air, and look out of the window a lot!

[nctnico]

Edit: I’ll just add ghat, as an aside, if there was universal centralised coordination model for road vehicles, road utilisation could be far higher and delays far lower. Practically speaking, it will be a long time coming, and not all road users will adopt it anyway.

This is being developed by various research institutes. I expect self driving vehicles will get some kind of mesh networking ability to coordinate between them and the road (traffic light, traffic density, etc) in order to use roads more effectively.