Autonomous Vehicles

[Amp34]

I didn't say remote control I said remote location. A Tonka truck moving about a very simplistic quarry is different level to a town center. If the technology is there to move along an empty road in northern Canada then why are they not doing it?

Where has town centre been mentioned anywhere (at least for lorries in the nearish future)? You’re making straw men again.

You said empty roads in northern Canada and I’ve been talking about main roads (such as motorways).

Why are these sort of vehicles not used on roads in northern Canada? Discounting the size and the fact most natural resources goods are transported by rail, legislation is probably a major factor. Legislation is something many western countries are looking at in detail right now.

Same with that guy who decapitated himself while watching harry potter in his back seat.

Ah yes, the incident where an HGV driver pulled dangerously across a main highway? Something that probably wouldn’t have happened if the vehicle was automated.

And to be clear, the technology in Tesla’s vehicles are level 2 (not the level 4 seen in the HGV demonstration or the taxis in Phoenix). It’s meant as an aid, but needs constant monitoring and the death was put down to driver error.

The National Transportation Safety Board has ruled that a truck driver’s failure to yield the right of way and a car driver’s “inattention due to over-reliance on vehicle automation” are the probable cause of the May 7, 2016, collision of a tractor-trailer and a Tesla Model S 70D sedan operating in autonomous mode.

Lots more info here from the National Transportation report on the matter- http://m.automotive-fleet.com/news/270597/ntsb-faults-both-trucker-and-motorist-in-fatal-tesla-crash

There’s also a question of whether the truck driver may have been impaired.

• Although the results of post-crash drug testing established that the truck driver had used marijuana before the crash, his level of impairment, if any, at the time of the crash could not be determined from the available evidence.

Our of interest do you think any automated vehicles won’t be around in the nearish future, or just Lorries?

 

[Joe T]

Well the more you practice something the better you get. Lots of hearings and investigations into crashes and near misses have pointed at the over use of assisted flying and lack of hands on flight time.

You're using your own logic rather than evidence.

The question is: is flight safety increased or decreased overall from the use of autopilot? That's the only question we need to answer.

 

[VincentHanna]

Again, by the time the Model 3 reaches us, it should be at least to level 4 automation.

That doesn't mean UK laws will enable you to use it of course, or that a lot of our roads would even be suitable.

 

[Amp34]

Just throw life into the mix and the self driving thing is flawed. Fuel on the road,

If it can be seen by a person then a camera can pick it up. Then advanced ABS and other systems go into affect.

traffic film covering a camera,

Already been sorted. Most semi autonomous vehicles don’t have an issue with film because they use things like infra red that can go through the film. Variables like rain, fog, hail and other weather problems that make it harder for humans to see are generally irrelevant to automated systems as it has little effect on what they can see.

cattle,

See Volvo’s large animal detection system as one example of this “problem” already being solved. Cattle is a much more minor issue than a deer or moose coming randomly out of a wooded roadside, which is what the tech is largely designed for.

high winds. So many variable situations.

Vehicles already have things like EBD and sway control for this sort of thing.

Most of the technologies to deal with those examples have been around for years/decades and doing so better than most humans ever could for much of it.

As an example I can go out this afternoon and pick up a vehicle that has sway control for both itself and a trailer it’s towing, then when I get to the destination I can get it to not only park itself, but reverse a large trailer into a parking spot too.

How do you teach AI to deal with a cow in the road or fallen tree? A blowout thats left the remains of the tyre sat 2 feet into the first lane of a motor way.

If there’s an obstruction a vehicle would either go round it, or if not possible just stop. It’s basically bread and butter automation you’re talking about here, automation that’s available on a significant amount of vehicles on sale right now (adaptive cruise control, large animal and pedestrian detection systems).

If it’s had a blowout, after using advanced EBD and ABS systems it slows itself down, pulls over and calls for a tow truck. Again it’s not magic. There are already plenty of vehicles that automatically call for assistance if they get into an incidence. Checkout Onstar, who are one company that provide this.

https://www.onstar.com/us/en/services/emergency.html

If you're in a crash, built-in sensors can automatically alert an OnStar Advisor and predict the severity of injuries. An Advisor is immediately connected into your vehicle to see if you need help, even if you can't ask for it. You don't have to do a thing.

For a blowout in particular the system could use the built in TMPS sensors in the tyres/wheels to gauge the issue alert the preferred authority.

Perhaps the biggest issue is that you don’t know how far advanced technology is right now?

 

[kinobestew123]

You're using your own logic rather than evidence.

The question is: is flight safety increased or decreased overall from the use of autopilot? That's the only question we need to answer.

It's not the only question, I asked one above. How is it going to affect people that enjoy driving? Just because it's statistically safer doesn't mean it's a done deal and should be implemented 100% with the manual option removed.

Shall we ban sport because it's safer not to play sport? How about ban leaving the house because it's safer to stay in the house?

 

[jigger]

You're using your own logic rather than evidence.

The question is: is flight safety increased or decreased overall from the use of autopilot? That's the only question we need to answer.

Google it. The more auto pilot is used the better it's record will be become. But if the system finds itself in a tricky situation it hands back control to the pilots to take the high failure risk. A problem that seems confounded by the dropping levels of experience.

 

[jigger]

If it can be seen by a person then a camera can pick it up. Then advanced ABS and other systems go into affect.



Already been sorted. Most semi autonomous vehicles don’t have an issue with film because they use things like infra red that can go through the film. Variables like rain, fog, hail and other weather problems that make it harder for humans to see are generally irrelevant to automated systems as it has little effect on what they can see.



See Volvo’s large animal detection system as one example of this “problem” already being solved. Cattle is a much more minor issue than a deer or moose coming randomly out of a wooded roadside, which is what the tech is largely designed for.



Vehicles already have things like EBD and sway control for this sort of thing.

Most of the technologies to deal with those examples have been around for years/decades and doing so better than most humans ever could for much of it.

As an example I can go out this afternoon and pick up a vehicle that has sway control for both itself and a trailer it’s towing, then when I get to the destination I can get it to not only park itself, but reverse a large trailer into a parking spot too.

Non of that seems like It's able to addresses examples of problems I listed.

 

[Mercenary Keyboard Warrior]

It's not the only question, I asked one above. How is it going to affect people that enjoy driving? Just because it's statistically safer doesn't mean it's a done deal and should be implemented 100% with the manual option removed.

Shall we ban sport because it's safer not to play sport? How about ban leaving the house because it's safer to stay in the house?

If, and assume it does, traffic incidents fall noticeably once we are running high levels of automation then we should probably remove human driving from the roads.
If there is indeed enough demand then racetracks are available (for goign quickly/pushing yourself) and lets face it, sims are getting so good then just get a decent pc/ xbox 2099 or whatever.

When you play sports you decide to put yourself in that situation, IMO, eventually when you go on the roads they will be derisked by the computing power deployed, we should probably not allow some wannabe Lewis H to dramatically increase risks just because he likes driving and is too cheap to go to a track. I see that as some 30+ years away or so though, as you need probably 20 years post automation to allow people to adapt and come round to its far easier to become a passenger than a driver.

 

[Amp34]

I never use my cruise control in heavy rain anyway, I'm not sure at the moment how good Tesla's are in the rain as it can confuse/obscure cameras anyway.

I wonder how the current (available) systems will react to snow on the road too. Presumably they can’t see the lane markings so will stop and/or prompt a driver to take over. It seems hard enough as a driver.

More advanced systems will know exactly where the road is so it’s unlikely to be a problem, and they may have sensors now that can physically see through the snow to the road below?

 

[Amp34]

You've said this a few times - does it actually affect passenger safety in the real world? Is there empirical evidence on that subject?

Ironically pilots may be becoming “worse”, but aviation safety continues to get better.

The issue is that when something catastrophic happens (such as that Air France accident over the Atlantic a few years ago) the pilots aren’t “experienced” enough to deal with it. The point being missed is that these incidents are becoming increasingly rare, which is why the pilots aren’t as experienced with dealing with them.

 

[Mercenary Keyboard Warrior]

I wonder how the current (available) systems will react to snow on the road too. Presumably they can’t see the lane markings so will stop and/or prompt a driver to take over. It seems hard enough as a driver.

More advanced systems will know exactly where the road is so it’s unlikely to be a problem, and they may have sensors now that can physically see through the snow to the road below?

For main roads they may need to add something to the lanes to give a better que for the systems.
Eg with my flooding issue, maybe they will develop smart sensors that can detect more than a certain amount of surrounding water and send a warning signal out. If they were linked they could talk down the line so you could be warned well in advance of upcoming issues. Nothing is impossible, just needs the will to do it. Obviously powered vs unpowered becomes the main issue.

Eg look at something like alaska they put poles down the sides of the roads so you can tell where the road is. Easy to add something similar to central reservations etc. High res mapping with detailed maps etc

 

[Amp34]

Non of that seems like It's able to addresses examples of problems I listed.

They’re specifically designed (and proven) to address the problems you listed.

 

[kinobestew123]

I wonder how the current (available) systems will react to snow on the road too. Presumably they can’t see the lane markings so will stop and/or prompt a driver to take over. It seems hard enough as a driver.

More advanced systems will know exactly where the road is so it’s unlikely to be a problem, and they may have sensors now that can physically see through the snow to the road below?

Presumably there would be no need for lane markings or lanes at all as cars could fit side by side with only a tiny gap and all move at any pace without colliding. Lanes are just a way of simplifying it for humans to comprehend.

The dream would be a system where every vehicle talked to ever other car on the road. If one sees an issue then every other vehicle knows about it instantly and can take evasive action. Vehicles just starting off on their route many miles away would automatically avoid the road.

Imagine a system where the vehicles are able to operate completely efficiently. No need for traffic lights as cars just seemlessly glide through junctions with mm to spare. No need to ever slow down, even in extremely busy areas. If a vehicle needs to come out of a junction, all vehicles slow down by the tiniest fraction many miles before thus creating the perfect gap for the vehicle to slide perfectly into.

 

[Nasher]

Already been sorted. Most semi autonomous vehicles don’t have an issue with film because they use things like infra red that can go through the film. Variables like rain, fog, hail and other weather problems that make it harder for humans to see are generally irrelevant to automated systems as it has little effect on what they can see.

Hmm no. You can't see through bad weather more clearly with IR. Maybe fog, but not rain or snow. Also LIDAR (which some of them use) really struggles with bad weather and there may be issues with cross-talk if lots of vehicles are using it in a small area.

But yea, any networked system is vulnerable to being hacked. That seems to be the problem they are ignoring.

 

[Amp34]

Hmm no. You can't see through bad weather more clearly with IR. Maybe fog, but not rain or snow. Also LIDAR (which some of them use) really struggles with bad weather and there may be issues with cross-talk if lots of vehicles are using it in a small area.

But yea, any networked system is vulnerable to being hacked. That seems to be the problem they are ignoring.

There are some issues with road film on some sensors, but most of the sensors can see at least as well as a human can. The solution to that problem is auto cleaning sensors, much like auto cleaning headlights now.

If everything is networked, youre right, there will be security concerns. This is a factor in much of modern day to day life. We deal with it by reducing the risk to an acceptable level. I really don’t see it as being as big an issue as some make it out to be. Much of our road system is already networked and subject to the same threats for example.

One thing to consider however is how they deal with different obstacles. A car hitting a balloon or plastic bag for example is not an issue, but a breeze block on the road is. Presumably it’s as simple as modelling the objects and training the system to “identify” the object in question, much like the animal detection systems.

 

[Simon]

I want to see 3 Self driving cars meet at a mini roundabout and see what happens.

Also if anyone knows Pangbourne. I want to see them drive through there. We already get complete gridlock due to people wanting to progress up the rod as far as possible. ( there are 3 bits with single car width, a roll bridge and loads of parked cars needing some creative driving at times.

 

[Nasher]

I want to see 3 Self driving cars meet at a mini roundabout and see what happens.

Also if anyone knows Pangbourne. I want to see them drive through there. We already get complete gridlock due to people wanting to progress up the rod as far as possible. ( there are 3 bits with single car width, a roll bridge and loads of parked cars needing some creative driving at times.

I'd like to see them do the "magic roundabout" in Swindon. They would be lost in there for months lol

There are some issues with road film on some sensors, but most of the sensors can see at least as well as a human can. The solution to that problem is auto cleaning sensors, much like auto cleaning headlights now.

If everything is networked, youre right, there will be security concerns. This is a factor in much of modern day to day life. We deal with it by reducing the risk to an acceptable level. I really don’t see it as being as big an issue as some make it out to be. Much of our road system is already networked and subject to the same threats for example.

One thing to consider however is how they deal with different obstacles. A car hitting a balloon or plastic bag for example is not an issue, but a breeze block on the road is. Presumably it’s as simple as modelling the objects and training the system to “identify” the object in question, much like the animal detection systems.

Yea until some joker throws a cement filled plastic bag in to the road, then you have a high speed pileup. So then you need a program to analyse movement as well and all this starts to get even more complicated and require more and more processing power, which means even more issues.

 

[Amp34]

That’s generally how similar systems work (pedestrian and animal detection), they analyze the shape AND movement. So if the cement filled bag isn’t wafting around, or rolling freely around the road looking like an empty plastic bag then it would be classed as an obstacle.

In other news Ford have 2021 set for their introduction and mass manufacture of level 4 autonomous vehicles for taxi and ride sharing applications.

https://corporate.ford.com/innovation/autonomous-2021.html

And the UK gov have just announced a similar timeline in the budget.

Driverless cars will be on Britain’s roads by 2021 as a result of sweeping regulatory reforms that will put the UK in the forefront of a post-Brexit technological revolution, Chancellor Philip Hammond will say this week. In his budget on Wednesday Hammond will allow driverless cars to be tested without any human operator inside or outside the car, and without the legal constraints and rules that apply in many other EU nations, and much of the US.

 

[jigger]

They’re specifically designed (and proven) to address the problems you listed.

No, they don't how do Volvo move a cow from a road?

 

[Amp34]

So now we’re not talking about driving safely, we’re talking about something in the vehicle getting out and waving their arms about?

TBH I’ve very rarely ever needed to get out of a vehicle and move an animal from the road and I have had to deal with animals on the road about a dozen times this year cows and deer mostly.

You’re right though, the current technology is not designed to shoo an animal away, it’s designed to stop a vehicle safely. It’s not going to be much extra effort to then program into the system driving round it slowly, the technology to drive around an obstacle is already available, along with the technology to identify the object in question.

Edit: and remember this is the technology available today on vehicles you can go out and buy. What is being developed in prototypes now, and what I’ll be available in 5 years time is going to be very different.