"These are completely safe autonomous vehicles": Cyclist spots driverless car using cycle lane

They will, if the AV parks where they want to park on the pavement / in the cycle lane...

Reading this reminded me of a trip a week ago when i was following a tradesman (a painter) up a local highway. A large tin of white paint had fallen over in his trailer and was leaking onto the road - Wheee, for several km two narrow lanes created from one normal lane. The autonomous car's AI system would have had fun with that. And then I got to thinking: Is this going to be the next method for vandalism? Teenagers painting white lines on the roads? e.g. bring an existing lane to a point. Or would this amount to terrorism if done on a high-speed freeway - imagine the potential chaos - paint in a few diagonal white lines and paint a few existing lines black.

It's an interesting article. I would however make two counterpoints when applying the logic to cars:

Firstly, pilots receive much more extensive and ongoing training and scrutiny than drivers. Given the behaviour of many drivers, there's a much lower bar for safety to be improved overall, even if automation introduces some new risks.

Secondly, and probably more importantly, cars have an obvious failsafe mode that planes don't: you can stop a car. If there's a problem that the computer can't deal with, all it needs to do is apply the brakes. If something goes wrong with a plane, it still needs to be able to keep flying and land safely. 

OnYerBike wrote:

Secondly, and probably more importantly, cars have an obvious failsafe mode that planes don't: you can stop a car. If there's a problem that the computer can't deal with, all it needs to do is apply the brakes. If something goes wrong with a plane, it still needs to be able to keep flying and land safely. 

Let us know how that works out for you when you do it in lane 3 on a busy motorway!

Brings me back to the first point: the only danger comes from other drivers who are driving carelessly. Automated cars aren't pefect but they are pretty good at not following too closely and not crashing into the back of the car in front when it slows down/stops.

I like the big sticker on the back of the car..."May stop quickly". Remind me, as I am sometimes very thick, but isn't that a consideration whenever following another mode of transport (even when walking)?

We'd like it to not crash into the back of other vehicles that might be on the road too.

And apparantly driving into a cycle lane regardless of who else might have thought they would remain on their part of the road. Good enough to avoid other cars does not equal good enough to be on the road.

Will this be the new standard for driving licenses? - Hey you drove around for 2 hours and didn't hit anything - great here's your licence.

Ashley Neal's description of his Tesla's assist: it's like a bad driver with good reactions.

They may not hit things if they can help it, but their aggressive and erratic manoeuvring can cause problems. You can often spot a driver planning to do something odd, would a Tesla give you the same cues?

The all need Rhonda Pickering as the passenger.

Failsafe mode on the flat maybe, I'd like to see AI handling a necessary stop on a 1:20 grade that has oil spilled on the road. Or where its necessary to keep on going no matter what, maybe on a road that is going under water because of rising floodwaters, or because of the approaching tornado, or the bushfire, or maybe your angry husband is chasing you with a gun. All rare and unusual incidents of course; but any study of local newspapers will show that rare and unusual is not so uncommon.

Although planes do need to keep flying, they start from a high altitude giving a lot of time to get back control. Whereas in a car, you may only have a few seconds to act which may be less than the time required to orientate yourself to the situation.

I'm not sure you can say the brakes can always be invoked if the system is controlling the car. I'm not that confident in bug free systems.

hirsute wrote:

... bug free systems.

No such thing. The goal is always to minimise bugs and try to ensure they do not compromise correct operation of the software. 

hirsute wrote:

Although planes do need to keep flying, they start from a high altitude giving a lot of time to get back control.

Unless the plane has only just taken off, as per both of the 737 Max crashes.

hirsute wrote:

I'm not sure you can say the brakes can always be invoked if the system is controlling the car. I'm not that confident in bug free systems.

That's a fair point, I never said it would be a perfect system. But the ability to "just stop", whether that be the car's reaction or a human intervention, is still there in a way that it isn't with an aircraft. 

OnYerBike wrote:

Unless the plane has only just taken off, as per both of the 737 Max crashes.

but that wasn't so much and aircraft failure as a failure of regulation and training, so very much like the tesla problems

Like this?

https://insideevs.com/news/630445/tesla-fsd-crash-san-francisco-footage/

I had actually seen that pop up on twitter - I maintain (as did many other commentors) that the crash was the result of the following drivers' careless/dangerous driving. It's not the UK so I don't know the letter of the law over there, but I'm pretty sure that the basic principle of leaving a sufficient distance to stop safely applies. 

They certainly contributed, but it does seem that the Tesla fail safe response was not very fail-safe. Surely simply removing drive would be a better response, akin to disengaging cruise control than actively stopping in a live motorway lane. If it was intelligent enough to be self-driving (albeit with supervision that it objected to detecting not being present), it should be bright enough to stop on the hard shoulder or exit the freeway and stop.

When would stopping on a motorway live lane ever be the response of first resort to an issue?

Bear in mind that the 2 second rule does not relate to a safe stopping distance.

Clearly stopping suddenly in a live lane is not ideal, and of course it shouldn't be a first resort. But as a last resort, it is (IMHO) more sensible than just cutting the power. And my original point was simply that it is a last resort option that is not available to an aircraft!

My understanding is that it was likely to be a response to a driver ignoring attention reminders. Whilst bringing the car to a halt may be an appropriate action in many urban environments, it is probably not the right choice on a freeway, where a moderate speed reduction and changing lane to the inside but progressing is safer than stopping. Given that the driver has already abrogated their responsibility, passing control to an inattentive driver on a high speed road does not seem well thought through. But that's the point isn't it? They are struggling to get self-driving cars to cope with normal scenarios, so abnormal scenarios are a big problem.

Sounds like drivers in Lancashire (thanks wtjs!).  Apparently no-one is providing them feedback either when they don't tax / insure their vehicles or drive them without due care and attention.  So maybe the autonomous car is being "trained" for operation there?

Sounds like drivers in Lancashire

I think I would still be safer around an AV 'autonomised' by a Raspberry Pi Zero than near Lancashire Audi and BMW drivers

Exactly.  We were supposed to have flying skateboards years ago, and we don't even have them yet, so 'properly' self-driving cars must be a long long way away...

OldRidgeback wrote:

... and also how to predict their behaviour.

Even though the point is that because cars deal with road conditions very differently to more vulnerable users (e.g. bicycles) cyclists can move in unpredictable ways.

The only solution is dedicated, segregated infrastructure.

For AVs.

kil0ran wrote:

The only solution is dedicated, segregated infrastructure.

For AVs.

Scalextric