Autonomous Vehicles

[Nasher]

Depends, if the manufacturer won't release telemetry from the AV (which they have refused to in the past) it might go the biker's way. They could say the "driver" should have checked their blind spots before moving lanes.

 

[Entai]

Can't comment on US law as I am not fully up to speed, so do not know how the manufacturer could refuse if they get a court order to release the data.

Certainly in the UK laws are being changed soon, (through the new Automated and Electric Vehicles Bill) so that after any accident or incident involving an AV, the manufacturer would be legally bound to release all data immediately, they would (by law) have to release all data to the insurance companies and to the police.

The new regulations and laws would require data from 30sec before and 15sec after the incident to be saved, to provide a clear answer as to what happened and which car is to blame, helping to speed up the insurance payout decision process, and any legal proceedings should any criminal act have been deemed to have occurred.

Also in the UK the new bill will state, that in an AV, in a situation as above, the "driver" is not in control, he would be technically, and legally, purely a passenger, while the AV is in control, until the moment the AV hands back control, which in a level 5 it never would obviously.

 

[sesevans]

Can't comment on US law as I am not fully up to speed, so do not know how the manufacturer could refuse if they get a court order to release the data.

......

I suppose in this case Nasher might be right about an OEM payout rather than a long, drawn out court case, unfortunately.

We seem to be in a political stalemate at the Federal level as the House of Representatives have passed self driving car regulation at the Federal level but the Senate has stalled on the bill. It will likely come attached to another bill this year rather than as the stand alone bill passed by the House.

So the current state of play is a patchwork of State laws (or in some case no laws), waiting for an eventual accomodation between State and Federal. In this sense, the UK legal framework seems to be ahead of the US.

https://www.bna.com/waymos-driverless-car-b73014474473/

 

[Chuk_Chuk]

Just some jumped up lawyer trying to make a name for himself I bet.

It could also be his insurance company. I was in a talk with an American lawyer and for some of these cases, the insurace company is the one to start the legal proceedings regardless of whether the person in question wishes to.

 

[Nasher]

Look like there was another one in California involving a Tesla again. Autopilot drove in to the back of a stationary fire engine at 65 mph

What they are saying here is quite interesting: https://www.wired.com/story/tesla-autopilot-why-crash-radar/

 

[jamoor]

My 2008 lexus does exactly this.

If youre using adaptive cruise your responsibility is then to make sure the computer is doing its job.

 

[sesevans]

Look like there was another one in California involving a Tesla again. Autopilot drove in to the back of a stationary fire engine at 65 mph

What they are saying here is quite interesting: https://www.wired.com/story/tesla-autopilot-why-crash-radar/

It is for these reasons that Waymo decided years ago to skip the idea of expecting the driver to intervene in any way in the AV. Their Level 4 robo-taxi service in Chandler, Arizona confirms this decision. They believe their robo-driver system must operate without any human intervention within the short time available to avoid a road traffic accident.

 

[Entai]

Look like there was another one in California involving a Tesla again. Autopilot drove in to the back of a stationary fire engine at 65 mph

What they are saying here is quite interesting: https://www.wired.com/story/tesla-autopilot-why-crash-radar/

Driver clearly not monitoring the situation constantly as it clearly says in the Tesla owners manual they should be, so purely the driver at fault and not the vehicle.


The issue here is that the Tesla is not really (in that manner) a true AV, as most of it's systems need constant monitoring, as it clearly says in the manual.

A true level 4 or level 5 AV, does not need to rely on constant monitoring by a human, so can adapt and can deal with situations like a parked vehicle easily.

 

[sesevans]

Waymo "Armada" being assembled, according to this link from Wired.

By the end of this year "thousands" of Waymo truly self driving vehicles will be operating in the 25 cities in the US in which they are currently testing small numbers of AVs. The US Federal Government is expected in 2018 to pass rules for AVs that could go as far as allowing AVs on US public roads without steering wheels and brake pedals, but Waymo is not waiting around for this to happen. It is deploying a huge number of AVs which do not require human intervention to operate because it believes its technology is ready.

https://www.wired.com/story/waymo-launches-self-driving-minivans-fiat-chrysler

 

[Nasher]

I'd put money on it all going wrong within a year after launch.

 

[sesevans]

I'd put money on it all going wrong within a year after launch.

I think you will have to define what you mean by "it all going wrong within a year".

Are you saying that Waymo for example, which could introduce many thousands of ride-hailing AVs operating in geo-fenced areas (ie fully 3D mapped with real time updates) in at least 25 US cities over the next year will be forced to pull the plug on the whole effort one year later? You are aware that Waymo has already clocked 4 million fully autonomously "driven" miles on various roads in various conditions and has clocked an additional 4 billion miles in virtual testing.

So what exactly do you mean and what amount of money are you prepared to "bet"?

 

[Nasher]

If it relies on a cloud system it's going to break at some point.

 

[Entai]

If it relies on a cloud system it's going to break at some point.

Hmmmmm

As we are already heading into the fourth year of cloud computing in many thousands of cars already on the roads of the world, do you want to change your statement at all?

It is well past a year and nothing has broken as yet.

 

[dowie]

You mean control like this:

Truck driver kills two pedestrians while playing Pokemon Go

Lorry driver kills four while on the phone

Romanian lorry driver who called two people was distracted by sat nav

Lorry driver kills van driver while checking how much time he had left

My commute down the M1 is often affected by accidents - the vast majority involve one truck or more.

I'm a firm believer that motorways need some form of automation.

frankly they should be banned form overtaking/forced to stick to the outside lane only and with a big gap between them and the truck or car in front... it seems to cause all sorts of knock on problems when you have one lorry driver very slowly overtaking another at like 1 mph faster

hopefully with automated lorries which can hopefully also be used/incentivised to be used more often at night then you'd see accidents fall as they can be set up to always stick in the outer lane, not overtake and always maintain a safe distance

Yes. Why not? We put up with humans driving them (approximately 90% of crashes are due to human driving by the way), so anything that makes our roads safer should be encouraged.

When Tesla added Autosteering into their range, there was a ~40% drop in the Tesla crash rate.

Tesla pulled off an awesome move when they got existing Tesla owners to essentially train their models, they've basically got self driving cars already.

 

[Nasher]

Hmmmmm

As we are already heading into the fourth year of cloud computing in many thousands of cars already on the roads of the world, do you want to change your statement at all?

It is well past a year and nothing has broken as yet.

A cloud system is a big single point of failure. Many companies have been bitten by this in the past few years and started bringing IT services back in-house. Just because it hasn't failed yet doesn't mean it won't spectacularly in future.

Also there have already been instances where GPS services have failed or developed issues (e.g. one was becoming a few meters out). What happens to all the AVs that rely on it for navigation?

 

[Entai]

A cloud system is a big single point of failure. Many companies have been bitten by this in the past few years and started bringing IT services back in-house. Just because it hasn't failed yet doesn't mean it won't spectacularly in future.

There have already been instances where GPS services have failed or developed issues (one was becoming a few meters out). What happens to all the AVs that rely on it?

If the guy who invented the wheel, thought "oh dear it might break one day I had better destroy my invention and never put it into use" we would all still be walking everywhere.


AV's do not rely on any one source of info, it is a huge network of information from dozens of sources and sensors, and if any one goes down there are backups of everything, so everything continues fine.

You do not grind to a halt at the side of the road, when you go under a bridge or into a tunnel and loose GPS or radio signal etc. do you ?

So why should an AV.

 

[sesevans]

.......

Tesla pulled off an awesome move when they got existing Tesla owners to essentially train their models, they've basically got self driving cars already.

I am still prepared to give Tesla the benefit of the doubt about their future growth but it is getting a little harder when they keep delaying delivery dates and performance metrics. For example, I recall they first said they would have their AV travel across the US by mid 2017, then by end 2017 and now, sometime in 2018. Considering how much data you say they have their on-road vehicles reporting, you would have thought they were well ahead of where they seem to be on AVs.

Meanwhile you have to appreciate the approach taken by Waymo, almost the complete opposite to Tesla. Waymo is not and does not want to be an vehicle manufacturer. Instead, like say Apple in smartphones, they select a manufacturer to take Apple's fully designed product and fabricate it according to Apple's specs. So Waymo/Fiat Chrysler seems a lot like Apple/Foxconn. Also Waymo has lined up other partners to help them deliver an on-road experience: Autonation for parts and repairs, Avis for maintenance, Lyft for additional options on ride sharing, etc.

Waymo appears to be moving towards launching a ride-hailing service in at least 25 US cities in 2018 having recently ordered thousands of AVs from Fiat. Each AV must carry a list price of $ 250,000 and an annual maintenace cost of $ 100,000. Hence Waymo is making a huge financial commitment to its ride hailing service. Clearly they believe their technology in their Level 4 vehicle is ready for prime time.

It is probably worth noting that the markets value Fiat Chrysler as more valuable a franchise now than Ford yet Ford has paid huge sums to buy autonomous technology to vertically integrate both manufacturing and software/AI. It will be interesting if they continue this approach. A big reorganisation plan is expected to be announced by Ford soon.

 

[sesevans]

There is a major trial pending in the US between Waymo and Uber over Uber's alleged theft of Waymo trade secrets. Uber has had a long history of legal troubles (Greyball, rescinding their license in London, etc) but this trial is probably the most significant they have ever faced because if the jury were to find them guilty of this theft, it could significantly slow down their ability to compete in a world moving towards AVs.

Jury selection begins today with the trial starting next week. Here is a link to the latest pre-trial arguments:

https://www.courthousenews.com/judge-winnows-what-jury-will-hear-in-waymo-uber-showdown/

 

[sesevans]

Meanwhile, while the Waymo v Uber trial proceeds to trial next week in the US, Waymo has both partnered with Lyft, Uber's US based rival, and invested in Lyft. Lyft opened its first ride sharing operation outside the US recently in Toronto and has begun work in Munich where, according to Bloomberg, "workers there will develop advanced localization and geometric mapping technologies to aid its autonomous driving effort."

https://www.bloomberg.com/news/arti...ves-into-munich-to-open-first-european-office

 

[Amp34]

Just reading an article published last onto about automated lorries and came across this.

A group at the Massachusetts Institute of Technology is gathering human input into the type of ethical dilemma such machines will face.

Participants are asked decide, for instance, whether a self-driving vehicle with brake failure should continue straight killing a woman, a baby, a criminal and a cat; or swerve, resulting in the death of a girl, a pregnant woman, a dog and a baby.

Now, in that situation what would you do?

Split second decision as you career towards them... what’s your answer. Especially pertinent to those who believe automated vehicles can’t make this sort of decision.

TBH I’d be pretty relieved to not have to make that decision and offload it to a computer if I actually came across a situation like that!