The Self-Driving Vehicles Thread

[Horse]

What I HAVE said - repeatedly - is that "you can't have a machine that appears to be in full control suddenly throwing its mechanical hands in the air" and expect a driver to take control.

Why do you assume that's when it will happen? Presumably you have a reason for that belief?

Tesla.

And you accuse me of being a stuck record?

Ok, have it your own way.

[MrLongbeard]

Tesla.

Pffft.
They can't even get their cars to be autonomous in a 75% closed system such as their own Las Vegas Loop.

[Horse]

Tesla.

Pffft.
They can't even get their cars to be autonomous in a 75% closed system such as their own Las Vegas Loop.

Didn't you know? It seems that Tesla are now the only people doing any development of these systems. Nothing else matters. In fact, it's going to be my word of the month for May.

Would you like a coffee? Tesla
Thanks for delivering that work early! Tesla

[MrLongbeard]

Didn't you know? It seems that Tesla are now the only people doing any development of these systems. Nothing else matters. In fact, it's going to be my word of the month for May.

Would you like a coffee? Tesla
Thanks for delivering that work early! Tesla

Well Musk has promised full level 5 cars by the end of this year.

[Horse]

Didn't you know? It seems that Tesla are now the only people doing any development of these systems. Nothing else matters. In fact, it's going to be my word of the month for May.

Would you like a coffee? Tesla
Thanks for delivering that work early! Tesla

Well Musk has promised full level 5 cars by the end of this year.

What's amazing is that he can get rockets to park more safely than the cars

[Cousin Jack]

So what if the driver isn't expected to take back control instantaneously? What if the system was designed to operate on its own for several - 'many' even - seconds. What it won't commit itself to a situation that it can't cope with?

The only way to be certain of "not committing itself to a situation that it can't cope with" is to refuse to drive. Period.

Unexpected situations are unexpected. And there are an infinite number of them too.

[Cousin Jack]

Well Musk has promised full level 5 cars by the end of this year.

And they will be quite safe - as long as they are confined to the US of A.

[Asian Boss]

I've seen Elon Musk's rockets landing in perfect synchrony and I've seen the gormless 17 year old boy next door having driving lessons. I'm fairly certain where the future lies.

[Horse]

So what if the driver isn't expected to take back control instantaneously? What if the system was designed to operate on its own for several - 'many' even - seconds. What it won't commit itself to a situation that it can't cope with?

The only way to be certain of "not committing itself to a situation that it can't cope with" is to refuse to drive. Period.

Indeed. As Spin will happily remind you (if he can stop himself from typing 'Tesla' ), there's no such thing as 'safe' [the MSF courses I taught in the 1990s had exactly the same point. Which means that, as you know , it's all about choosing an acceptable level of risk.

Funnily enough, there have been people working on whether an increase in risk might be temporarily acceptable. The only example I can remember from the presentation (heard about it years ago) was the AV travelling in lane 1 moving into lane 2 to allow a merging vehicle room - but having to fit between two other vehicles already in lane 2 temporarily resulting in acceptance of shorter headway.

Unexpected situations are unexpected. And there are an infinite number of them too.

Yes, but ...

I don't know about cars, but let's look at bikes. Riders typically have three main 'gotcha' crashes:
- 'right of way violations', car turns out or across
- bends, loss of control, usually running wide
- overtaking (including filtering as a sub-set)

If you want more detailed figures, they're available.

And, if you watch that video I posted, you'll see that some of the work being done is to use crash investigation information to test the AV software before it gets anywhere near the road. Not just that, the virtual testing allows for umpteen variations to be tested. Using crash information for learning, something the 'Safety II' team will emphasise as good practice. And FWIW I was using a 'top ten' breakdown of crashes in my 1988 advanced course theory session.

So 'infinite variation' perhaps. But if that was totally the case then it would scupper Spin's 'No Surprise' campaign. Which, of course, it doesn't, because the main elements are fairly easy to identify. NS even has its rhyming reminders, like 'gaps = traps' & 'can go = will go'. If a person can learn them, then a machine probably can too.

The only way to be certain of "not committing itself to a situation that it can't cope with" is to refuse to drive.

I recently heard someone say about their AV: "the problem wasn't stopping it, the problem was keeping it going". Yup, the system was so 'careful' that it would stop at just a hint of trouble.

[demographic]

I've seen Elon Musk's rockets landing in perfect synchrony

Yeah, I've seen footage of them doing a rapid unscheduled disassembly on a number of occasions as well.

[Asian Boss]

I've seen Elon Musk's rockets landing in perfect synchrony

Yeah, I've seen footage of them doing a rapid unscheduled disassembly on a number of occasions as well.

Ha, yes. Me too. It does go badly wrong for them sometimes still. But the computer stuff will get better, possibly an an exponential rate. People will not.

[Cousin Jack]

Ha, yes. Me too. It does go badly wrong for them sometimes still. But the computer stuff will get better, possibly an an exponential rate. People will not.

Once upon a time we couldn't be trusted to drive above 4 mph and with a man with a red flag walking in front.
Some of us now get up to 70 mph without problems.

Might not be exponential, but we are getting better.

[wheelnut]

I've seen Elon Musk's rockets landing in perfect synchrony and I've seen the gormless 17 year old boy next door having driving lessons. I'm fairly certain where the future lies.

The problem is, the process of driving a car is a lot more chaotic and harder to predict than the act of flying a rocket (or a plane for that matter).

[Horse]

Ha, yes. Me too. It does go badly wrong for them sometimes still. But the computer stuff will get better, possibly an an exponential rate. People will not.

Once upon a time we couldn't be trusted to drive above 4 mph and with a man with a red flag walking in front.
Some of us now get up to 70 mph without problems.

Might not be exponential, but we are getting better.

Have humans improved (how?), or have vehicles and infrastructure, and medical care, been improved to cater for continued and unchanged human weaknesses?

The consequences of a 70 mph fall from a motorcycle (in terms of the energy to dissipate and potential for injury) are no different now from 1896 when the red flag was repealed.

Has human perception or performance (reaction time, etc) improved since 1896?

[Horse]

The problem is, the process of driving a car is a lot more chaotic and harder to predict than the act of flying a rocket (or a plane for that matter).

Many people use the 'downstairs loo' as a gallery. A friend has a framed photo in his, showing him surrounded by his team, posing in front of a large trailer which is carrying his rocket. He says that, contrary to the old saying, rocket science is easy

Re chaotic. Yes-ish. See previous answer to CJ.

[Horse]

Ok, I'm going to say this once, see what happens ...

Tesla




Well, the world hasn't imploded.

In the news today: remember the 'no-one in the driving seat' crash. Well, according to Tesla:

preliminary data downloaded by Tesla indicate the vehicle was not operating on Autopilot, and was not part of the automaker's "Full Self-Driving" (FSD) system.[/]

Perhaps the occupants really were desperate for that Darwin Award?


Also, I had wondered how the US 'views' such vehicles. Apparently:

U.S. federal road safety authority, the National Highway Traffic Safety Administration (NHTSA), has yet to issue specific regulations or performance standards for semi-autonomous systems such as Autopilot, or fully autonomous vehicles (AVs).

There are no NHTSA rules requiring carmakers to ensure systems are used as intended or to stop drivers misusing them. The only significant federal limitation is that vehicles have steering wheels and human controls required under federal rules.

https://www.reuters.com/technology/tesl ... 021-04-20/

[The Spin Doctor]

Well, according to Tesla:

According to Boeing, there was nothing wrong with the 737 Max, either, when the two planes crashed.

Let's wait for an INDEPENDENT analysis first.

preliminary data downloaded by Tesla indicate the vehicle was not operating on Autopilot, and was not part of the automaker's "Full Self-Driving" (FSD) system.[/]

This is how Tesla sells 'Autopilot' which is NOT fully autonomous.

Tesla Autopilot.JPG (40.5 KiB) Viewed 790 times

[Horse]

Well, according to Tesla:

According to Boeing, there was nothing wrong with the 737 Max, either, when the two planes crashed.

Let's wait for an INDEPENDENT analysis first.

Indeed. Exactly why I stated, and included in the quote, that it was a statement from Tesla.

You have actually stumbled into a bigger issue, of how manufacturers support (or obstruct) data downloads.

And many riders might be surprised by how much is being logged - and available post-crash - by their bikes.

[Cousin Jack]

You have actually stumbled into a bigger issue, of how manufacturers support (or obstruct) data downloads.

And many riders might be surprised by how much is being logged - and available post-crash - by their bikes.

Which may be a blessing in disguise. If all the data logged was available in a nice standard format you would get penalty points for speeding EVERY time you went over a limit. At the moment it may be available, but it is a PITA to recover, so only a serious crash makes it worthwhile.

[Horse]

Well, again from a limited understanding, the ease of access is varied.

But are HGV tachograph logs used for routine speed prosecutions etc.? They're about as easy to access as it could be.

Don't replace your Arai with a tinfoil hat quite yet