And you thought these driverless cars were scary!

[flying dog]

Someone just sent me this link....

 

 

[Lucky01]

Wow.

[Old Koreelah]

Didn’t the Brits have autoland on the 1960s Trident?

 

In a few years pilots might be lumped together with blacksmiths and stenographers.

[turboplanner]

You can buy the technology for return to base and autoland in a $60.00 drone today.

As this shows, the algorithms can be produced for all the adjustments to controls, engine settings and possibly even identifying duty runways and Tower commands, a Mayday can be relayed etc.

 

The problem comes with writing one to cope with every mistake that every pilot in the history of flying has even made, and every failure of every component an aircraft has suffered and every known fix; that gets into the trillions.

[Old Koreelah]

1 hour ago, turboplanner said:

The problem comes with writing one to cope with every mistake that every pilot in the history of flying has even made, and every failure of every component an aircraft has suffered and every known fix; that gets into the trillions.

Instead of designing for umpteen possible human stuff-ups, they’ll most likely design for total autopilot, with optional human input (on the proviso that the human plays nice).

[onetrack]

How hard must it be, for that bloke sitting in the front left seat, to resist grabbing any of the controls, as it touches down! I noticed his hand creeping forward to grab the joystick, as it hit the runway! 

[Old Koreelah]

1 minute ago, onetrack said:

How hard must it be, for that bloke sitting in the front left seat, to resist grabbing any of the controls, as it touches down! I noticed his hand creeping forward to grab the joystick, as it hit the runway! 

Nothing that couldn’t be fixed by a robot clamping his wrists!

 

https://jalopnik.com/the-thought-of-a-single-pilot-airliner-is-just-a-bit-sc-1677413476

[Ian]

Total automation of the flight envelope would probably be both safer and cheaper however people don't make decisions based on logic. That's not to say that there wouldn't be incidents however from a computing perspective flying and landing isn't that difficult. Especially while you have an operational GPS system.

However systems need to be designed effectively with the dumb design decisions associated with the boeing max where a single point of failure model was seen as appropriate. Essentially this was a half arsed attempt at automation. Airbus used triple reduncancy for the same sensors.

 

https://www.faa.gov/sites/faa.gov/files/about/office_org/headquarters_offices/avs/CFIT_brochure.pdf

From the above the top two in GA are

• Loss of control in flight
• Controlled flight into terrain

Interestingly commercial aviation fatalities shows the same no 1 and 2

https://flightsafety.org/wp-content/uploads/2017/07/CFIT-Report-1st-Ed-2015.pdf

 

Another  fact is that controlled flight into terrain has reduced markedly over the past few years however controlled flight into terrain has remains significant.

https://www.iata.org/contentassets/b6eb2adc248c484192101edd1ed36015/loc-i_2019.pdf

 

[facthunter]

Watching the automatics doing the Landing is not much different to watching the Other pilot do it, especially when training with assymetric power, where  it's easier to go pear shaped..  Nev

[turboplanner]

The aircraft is easy; the environment it works in is not.

 

[sfGnome]

10 hours ago, Ian said:

Total automation of the flight envelope would probably be both safer and cheaper however people don't make decisions based on logic.

Elon Musk made an interesting comment regarding full automation in motor vehicles. Paraphrasing, he said that getting self drive cars up to the same standard as human drivers was setting the bar very low. Everyone expects other drivers to make mistakes, but fully automated ones are expected to never do so.

[kgwilson]

14 hours ago, Old Koreelah said:

Didn’t the Brits have autoland on the 1960s Trident?

 

In a few years pilots might be lumped together with blacksmiths and stenographers.

The Hawker Siddeley Trident was originally designed by Geoffrey De Havilland as the DH-121 in the mid 1950s. It was designed from the outset to autoland so that fog which was very common then due to a lot of coal fires etc would not prevent landings. It was the first by a country mile and 70 years. De Havilland was a genius ahead of his time having designed the most versatile aircraft in WW2 the Mosquito and out of wood to boot as well as the first ever Jet airliner, the Comet.

[flying dog]

8 hours ago, sfGnome said:

Everyone expects other drivers to make mistakes, but fully automated ones are expected to never do so.

 

Alas I fear that is a big problem / flaw with that idea.   It is wrong.

If another self driving car is near by and *something goes wrong*.....   All bets are off.

 

(BTW, the signature of that post was very good.)

[Ian]

Self driving

1 hour ago, flying dog said:

Alas I fear that is a big problem / flaw with that idea.   It is wrong.

If another self driving car is near by and *something goes wrong*.....   All bets are off.

I don't really follow your logic here and I not sure what you are trying to say?

Autonomous cars drive by themselves, so if another car nearby goes haywire it will deal with them in the same manner as one of the many thousands of nuts currently behind the wheel. All bets are still on because they're autonomous they just react to another crazy driver. It's a little different being the occupant of the vehicle going crazy though but then you're in the same situation as taking a lift with someone and realizing that they're a pretty crap driver.

 

Or were you trying to make a different point?

 

Any moving powered object has a temporal bubble which indicates it's possible locations at some future point in time, the greater the time the larger the bubble. It's just maths and number crunching, some of these states are unlikely and that's what the systems try to work with on a crowded road. However the temporal bubbles can only be projected fractions of a second ahead without intersection.

With planes it's much easier as there's so much more room that you can largely stay outside these temporal bubbles even when they're projected minutes ahead.

 

There's a few stats published by tesla.  https://www.tesla.com/VehicleSafetyReport

In the 4th quarter, we recorded one crash for every 4.31 million miles driven in which drivers were using Autopilot technology (Autosteer and active safety features). For drivers who were not using Autopilot technology (no Autosteer and active safety features), we recorded one crash for every 1.59 million miles driven. By comparison, NHTSA’s most recent data shows that in the United States there is an automobile crash every 484,000 miles.

 

 

 

[flying dog]

I read the post as "Autonomous cars drive with the idea that every car drives safely."

 

> but fully automated ones are expected to never do so.

 

That is a good idea/theory, and would be valid most of the time.

 

But if one of these cars went crazy and wasn't driving safely, then YOUR self driving car would have to deal with a car which wasn't obeying the rule just stated.  (All cars drive safely.)

And so it would be confronted with an unknown scenario.

 

Edited June 13, 2022 by flying dog
Adding the quote for refrence

[onetrack]

Every automated driving system is full of fail-safe protections. In mining today, sizeable fleets of dump trucks, and even the excavators that load them, are automated.

Automated mining equipment was introduced in the mid-1980's, it was operated like RC models in those days, where there was extreme danger to the operator.

 

Today, the mining equipment is fully automated, and merely overseen from remote positions, up to even 2000 or 3000 kms away.

If the sensors and cameras pick up anything even slightly wrong, the machine stops immediately - and the other machines stop as well, because they all work together.

The incidence of accidents and damage is substantially lowered with automated mining equipment, as compared to careless human operators, human operators lack of attention, human operators slow response to danger developing.

 

I have little doubt that driverless cars will be with us on a large scale within 20 years. The major restriction at this point is the lack of standardised infrastructure, signage, and in-road sensors. They will all come in time.

[sfGnome]

2 hours ago, flying dog said:

I read the post as "Autonomous cars drive with the idea that every car drives safely.".

 

Sorry, I think that was a poor paraphrase on my part. I was meaning that the autonomous car is expected to handle any craziness around it without ever resulting in any accidents.