Will robots/AI replace human pilots?

We've had autoland for a long time; C-141s had it more than half a century ago. Probably some other, perhaps non-production aircraft had it earlier?

Art and science of software development is still kinda stoneage; there's plenty of applied science cool stuff to do within the current methods, but until there is a real breakthrough in software engineering, one that can compete and persist, the state of the art is too crude to build truly autonomous systems at prices that make sense. They'll need "overseers" for a long time to come.
 
As far as you know.

Nauga,
who sees mistrust run deep

Well rephrase it. Very few airplane engineers running screaming to stop their extended families from boarding these days. Haha.

It's not the 1930s when people had to be enticed on board with fancy meals and pretty girls, that's for sure.

Here's your peanuts (if you're lucky) and don't talk to the FA. Just don't. Hahahaha.

It'll be "fun" to see how they market the first passenger drone flights. Lemme guess... a fine meal and pretty girls and guys in those FA uniforms. (Well, assuming the things have FAs...)

The pretty girls thing didn't work out for Icon but nobody can say they didn't try. :)
 
We've had autoland for a long time; C-141s had it more than half a century ago. Probably some other, perhaps non-production aircraft had it earlier?
Just finished a four-day, six-leg trip in 737NGs which all have autoland capability. We didn't do any autolands but, even if we had wanted to, only half of the six legs were eligible. Two of them had winds out of limits for autoland and the third was at an airport without a useable runway that would support autoland.

Autoland allows us to land in conditions where there isn't enough visibility to land manually but it comes with a lot of restrictions, including wind limits, which are frequently exceeded in normal operations. With very low visibility (<1800 RVR) there's rarely much, if any, wind so it typically isn't a problem though you still need a runway and approach that will support it.

The last landing on the trip was with a 32kt direct crosswind (24G32 IIRC). That's more than double the autoland's crosswind limitation. Even if the wind had been aligned with the runway it would have exceeded the autoland's headwind limitation.
 
One statement - there's a huge difference between being able to get it right and not being able to get it wrong.
 
Here's an interesting TED Talk which discusses the state of AI. I think a lot of it is applicable to the autonomous airliner.

 
I have not read the thread. I am only responding to the title.

I am a retired automation engineer. I have automated everything from pipelines to paper mills to steel mills to food preparation lines and on and on and on. There are many things that lend themthelsves to automation and some not.

SO... I ask you, do you think a robot could have made the decision to do so and then successfully ditched an airliner in the Hudson and walked the cabin to ensure everyone was out safely?
 
Yes

Btw..... these thingys aren't run with PLC code.
 
Last edited:
I have not read the thread. I am only responding to the title.

I am a retired automation engineer. I have automated everything from pipelines to paper mills to steel mills to food preparation lines and on and on and on. There are many things that lend themthelsves to automation and some not.

SO... I ask you, do you think a robot could have made the decision to do so and then successfully ditched an airliner in the Hudson and walked the cabin to ensure everyone was out safely?

I think that a robot could analyze the parameters of a dual engine failure in a split second and would have turned back to LGA and saved the entire crew, passengers and aircraft.:eek:
 
Pilot Tells of Moment 'Psycho' Flight Computer Made Plane Nosedive 1,000 Feet (The Telegraph, May 15)
219ddfaa-710d-47eb-9ce0-9477aad3a0f0.jpg
A pilot who saved the lives of more than 300 people by averting the crash of his passenger jet has broken his silence after nine years, describing how the aircraft's computer "went psycho" plunging the plane hundreds of feet in seconds.
Captain Kevin Sullivan was in charge of Qantas Flight 72, from Singapore to Perth, when the Airbus A330's auto-pilot malfunctioned twice, first dropping the aircraft 690 feet in 23 seconds, and again sending it 400 feet in 15 seconds, causing more than 100 injuries to passengers.
On that day in October 2008, Sullivan was forced to declare mayday and make an emergency landing at Learmouth Airport in Western Australia, gripped with the fear that the plane might nosedive again.
"It's the worst thing that can happen when you are in an aeroplane - when you are not in control," he told Good Weekend, part of the Sydney Morning Herald.
"And you have a choice. You can either succumb to that or you fight it. I was fighting that outcome and I have been ever since."
Flight 72, carrying 303 passengers and 12 crew, had been cruising at 37,000 feet above the Indian Ocean when one of the plane's three flight control primary computers - which pilots know as Prims - developed a fault and warnings of both stall and over-speed began ringing around the flight deck. Moments later the A330's nose pitched down violently, sending chaos through the cabin.
Sullivan, a former US Navy pilot, said he asked himself "Is my life going to end here today?" before managing to pull the plane level again using the control stick.
"We're in an out-of-control aeroplane, we're all juiced up by our own bodies because, we thought, we are in a near-death situation, and we've got to be rocket scientists to figure out how we can go in there and land the plane outside of any established procedures," he said.
Sullivan says the experience, which left him with post-traumatic stress disorder, raises questions about the increased automation of flying.
"We were never given any hint during our conversion course to fly this aeroplane that this could happen," he said. "And even, I think the manufacturer felt this could never happen. It's not their intention to build an aeroplane that is going to go completely haywire and try and kill you."
"Even though these planes are super-safe and they're so easy to fly, when they fail they are presenting pilots with situations that are confusing and potentially outside their realms to recover," he told the Sydney Morning Herald. "For pilots - to me - it's leading you down the garden path to say 'You don't need to know how to fly anymore'. You just sit there - until things go wrong."
Sullivan, an American who moved to Australia for what was meant to be a three-year stint but stayed after marrying an Australian, kept working for Qantas for eight years, remaining quiet on the incident, until last year when he left the airline.
It was just 50 minutes after the plane's first nose-dive that Sullivan and his two co-pilots had the plane on the tarmac at Learmouth Airport, where they were met by emergency services.
Three years later, a report by the Australian Transport Safety Bureau found that incorrect data sent to one of the aircraft's computers had caused the nose-dives as the plane sought to correct perceived errors that did not actually exist.
https://goo.gl/rQsN61
 
Yes

Btw..... these thingys aren't run with PLC code.

I fully realize they are not run with PLC code. I also know that the answer to my question is no. In 99.9% of situations it could fly pilotless. Deciding to take the river and have the stick and rudder feel to ditch successfully. Don't rely on it.
 
In addendum to the last post.

MD11 approximately 700 nm off coast of Ireland, all three engines went to idle and a few seconds later back to normal speed and then back to idle. Many warnings, oral and lights. Three of us in the cockpit, manuals out (there is no checklist for this) company maintenance on SATCOM and feeding through to Boeing Tech whose first words were "that can't happen"...well it is, dig deeper. Basically the fuel sensors were telling the FMC that we had lost 50000 lbs of fuel then it would come back. Airplane was flying and we did what pilots do...shut off all the automation and flew like pilots. Turns out that the fuel transducers in the tanks send signals to a processor that send it to the main computers. System designed so that the tank transducers swap out signals to give a redundancy but if one is bad then it sends bad data down the line. Simple, just disconnect the bad one! How? No procedure, no tests to determine from the cockpit, Boeing was clueless also because "it can't happen".
Crew was blames for diverting to a non company airport. Their EXPERT maintenance guy flown in said we had plenty of fuel so he dipped the tanks and using modern math determined that we had more fuel onboard than we took off with! We went to the hotel.
 
...Their EXPERT maintenance guy flown in said we had plenty of fuel so he dipped the tanks and using modern math determined that we had more fuel onboard than we took off with! We went to the hotel.
It's a MIRACLE! ;)
 
Last edited:
...company maintenance on SATCOM and feeding through to Boeing Tech whose first words were "that can't happen"...

We installed the conference bridge that handled those calls in the 90s. Heh. I'm sure it's long-retired by now. Some of the "phone calls" that were "monitored by support technicians for audio quality" ahem, the airplane geek of the install and support team anyway, cough, me... were rather entertaining. Eye opening too. :)

Let's see. When did Boeing gather up MD? That was probably before then. The new hotness back then was probably the 737-300 and the 757. Too long ago. I'd have to go look at Wikipedia. Those brain cells were culled to keep the herd sharp via beer long ago. Ha.

Was always impressive how good y'all sounded over the Sat links. The Airinc links had level problems, depending on the site used. Airinc didn't understand the whole -7/-12 dB thing when feeding a T1 from old 4-wire analog gear back then.
 
Yeah, and Sully made the right decision.

The simulations showed the plane could have landed but the scenario failed when adding in the delay time Sully had when he made the decision to fly out and land on the river. A machine would have made those calcs quicker and landed back on the runway.
 
A machine would have made those calcs quicker and landed back on the runway.
You're using 20/20 hindsight which not even your fictional machine would have had access to at the time of the bird ingestions. You're also overestimating the speed at which the calculations could be made.
 
For every Qantas incident that COULD have killed people, how many incidents DID kill people as a result of human error?

There's your answer as far as automation it concerned. If a computer kills people .00001% of the time and a human kills people .00002% of the time, the computer is safer. Period. Full stop. End. Fin.

Whether you can mentally accept a computer killing people instead of demanding that a person kill people is the real issue at hand.
 
Last edited:
Pilot Tells of Moment 'Psycho' Flight Computer Made Plane Nosedive 1,000 Feet (The Telegraph, May 15)
219ddfaa-710d-47eb-9ce0-9477aad3a0f0.jpg
A pilot who saved the lives of more than 300 people by averting the crash of his passenger jet has broken his silence after nine years, describing how the aircraft's computer "went psycho" plunging the plane hundreds of feet in seconds.
Captain Kevin Sullivan was in charge of Qantas Flight 72, from Singapore to Perth, when the Airbus A330's auto-pilot malfunctioned twice, first dropping the aircraft 690 feet in 23 seconds, and again sending it 400 feet in 15 seconds, causing more than 100 injuries to passengers.
On that day in October 2008, Sullivan was forced to declare mayday and make an emergency landing at Learmouth Airport in Western Australia, gripped with the fear that the plane might nosedive again.
"It's the worst thing that can happen when you are in an aeroplane - when you are not in control," he told Good Weekend, part of the Sydney Morning Herald.
"And you have a choice. You can either succumb to that or you fight it. I was fighting that outcome and I have been ever since."
Flight 72, carrying 303 passengers and 12 crew, had been cruising at 37,000 feet above the Indian Ocean when one of the plane's three flight control primary computers - which pilots know as Prims - developed a fault and warnings of both stall and over-speed began ringing around the flight deck. Moments later the A330's nose pitched down violently, sending chaos through the cabin.
Sullivan, a former US Navy pilot, said he asked himself "Is my life going to end here today?" before managing to pull the plane level again using the control stick.
"We're in an out-of-control aeroplane, we're all juiced up by our own bodies because, we thought, we are in a near-death situation, and we've got to be rocket scientists to figure out how we can go in there and land the plane outside of any established procedures," he said.
Sullivan says the experience, which left him with post-traumatic stress disorder, raises questions about the increased automation of flying.
"We were never given any hint during our conversion course to fly this aeroplane that this could happen," he said. "And even, I think the manufacturer felt this could never happen. It's not their intention to build an aeroplane that is going to go completely haywire and try and kill you."
"Even though these planes are super-safe and they're so easy to fly, when they fail they are presenting pilots with situations that are confusing and potentially outside their realms to recover," he told the Sydney Morning Herald. "For pilots - to me - it's leading you down the garden path to say 'You don't need to know how to fly anymore'. You just sit there - until things go wrong."
Sullivan, an American who moved to Australia for what was meant to be a three-year stint but stayed after marrying an Australian, kept working for Qantas for eight years, remaining quiet on the incident, until last year when he left the airline.
It was just 50 minutes after the plane's first nose-dive that Sullivan and his two co-pilots had the plane on the tarmac at Learmouth Airport, where they were met by emergency services.
Three years later, a report by the Australian Transport Safety Bureau found that incorrect data sent to one of the aircraft's computers had caused the nose-dives as the plane sought to correct perceived errors that did not actually exist.
https://goo.gl/rQsN61

Wait - The pilot was a former US Navy pilot, and he was quoted using the term "aeroplane?" Hmmm....
 
You're using 20/20 hindsight which not even your fictional machine would have had access to at the time of the bird ingestions. You're also overestimating the speed at which the calculations could be made.

Fore Flight has a Glide Advisor now that shows a boundary around the aircraft in glide distance. If a simple App like that can project where the aircraft can glide to, I'm sure we can come up with software in an airliner similar to that. Combine it with an AP and program it to automatically go for nearest airport.
 
Another thing about the Qantas flight... The number it gives are for a 1200 fpm dive... I routinely do that when I suck at calculating my descents. I guess it's just a matter of the rapidity of the control input?
 
Fore Flight has a Glide Advisor now that shows a boundary around the aircraft in glide distance. If a simple App like that can project where the aircraft can glide to, I'm sure we can come up with software in an airliner similar to that. Combine it with an AP and program it to automatically go for nearest airport.
a lil bit more than that.....but conceptually you have it. :)
 
Fore Flight has a Glide Advisor now that shows a boundary around the aircraft in glide distance. If a simple App like that can project where the aircraft can glide to, I'm sure we can come up with software in an airliner similar to that. Combine it with an AP and program it to automatically go for nearest airport.
Calculating a glide radius in not the problem, though it won't know how the wind will change during the descent and calculating a landing pattern is a lot more difficult than a radius.

You're using 20/20 hindsight to assume that the system will immediately conclude that the two engines are completely failed (they weren't, BTW) and that they can not be restarted. The river gave flexibility to absorb the effects of the unknown factors. The short runways at LGA did not.

An automated airliner has to solve these problems without already knowing the outcome. It has to solve the next problem that none of its programmers ever imagined. See the video in post #205.
 
Fore Flight has a Glide Advisor now that shows a boundary around the aircraft in glide distance. If a simple App like that can project where the aircraft can glide to, I'm sure we can come up with software in an airliner similar to that. Combine it with an AP and program it to automatically go for nearest airport.
Have you tested the Glide Advisor in real conditions?

It's rather obvious it doesn't include the effect of winds. Only the input glide ratio and terrain.

It amazes me just how bad the armchair engineering can be on this board.

For the guys who think a machine "could have" landed at LGA, point out the procedures for a double flameout in the appropriate AFM. The multitude of verbiage on that topic indicates how much design would have gone into that problem before it happened.

Technology is not magic. Someone has to design and test the response. No one did, so its odds of working were ZERO.
 
Have you tested the Glide Advisor in real conditions?

It's rather obvious it doesn't include the effect of winds. Only the input glide ratio and terrain.

It amazes me just how bad the armchair engineering can be on this board.

For the guys who think a machine "could have" landed at LGA, point out the procedures for a double flameout in the appropriate AFM. The multitude of verbiage on that topic indicates how much design would have gone into that problem before it happened.

Technology is not magic. Someone has to design and test the response. No one did, so its odds of working were ZERO.
Foreflight can't include winds because it doesn't have an input for the winds on your aircraft. If it did, it could. The G1000 I flew knew which direction the winds were blowing and how hard they were blowing for each and every foot of whatever altitude I was at. If foreflight had that data being plugged into it, it could easily figure out glide.

Technology is and has been creeping into every single aspect of our lives since technology was a thing. We get around using metal cages sitting on spinning metal attached to tree syrup, which in turn are powered by more spinning metal being spun by burning dead dinosaur juice. We communicate with each other across the globe at the speed of light by a series of tubes. We FRIGGIN FLY. How much of a Luddite does one have to be to not realize that automation of as many things as possible is going to be a very real thing in the future. Maybe not tomorrow or 5 or 10 years from now, but it absolutely WILL come to fruition.
 
Have you tested the Glide Advisor in real conditions?

It's rather obvious it doesn't include the effect of winds. Only the input glide ratio and terrain.

It amazes me just how bad the armchair engineering can be on this board.

For the guys who think a machine "could have" landed at LGA, point out the procedures for a double flameout in the appropriate AFM. The multitude of verbiage on that topic indicates how much design would have gone into that problem before it happened.

Technology is not magic. Someone has to design and test the response. No one did, so its odds of working were ZERO.

With a winds aloft input from XM or ADS-B, it will vary the glide boundary.
 
With a winds aloft input from XM or ADS-B, it will vary the glide boundary.
why not just us actuals?...no need to interpolate with forecast data.

Observe ground speed (GPS)....with an optimized glide speed (TAS)....use that to calculate the DZ. Iterate until the impact site is reached. :D
 
why not just us actuals?...no need to interpolate with forecast data.

Observe ground speed (GPS)....with an optimized glide speed (TAS)....use that to calculate the DZ. Iterate until the impact site is reached. :D

And if the wind is different below (and it usually is), you have the wrong answer.

You need to know the INTEGRATED winds at all altitudes below you, along your trajectory. There is no data source for that. One could imagine doing that with a nonexistent directional narrow-band Doppler radar carefully tuned to a water absorption line, but there is a whole helluva lot of engineering to design such a beast.

It's not good engineering to assume that data is good. You must consider its source, and its meaning. An XM forecast of winds aloft is only slightly better than having no winds at all.
 
And if the wind is different below (and it usually is), you have the wrong answer.

You need to know the INTEGRATED winds at all altitudes below you, along your trajectory. There is no data source for that. One could imagine doing that with a nonexistent directional narrow-band Doppler radar carefully tuned to a water absorption line, but there is a whole helluva lot of engineering to design such a beast.
Meh....it doesn't change that much. My answer sez....it dun matter.
 
Meh....it doesn't change that much. My answer sez....it dun matter.

No, it doesn't.

You have to pick your landing zone at the top of glide, while you still have options. Your analysis presumes you have done that.

Classical numerical fallacy assuming all minima can be found by local means. They can't. Your best answer is NOT to follow the local wind, but to locate a good landing zone.
 
No, it doesn't.

You have to pick your landing zone at the top of glide, while you still have options. Your analysis presumes you have done that.

Classical numerical fallacy assuming all minima can be found by local means. They can't. Your best answer is NOT to follow the local wind, but to locate a good landing zone.
you pick your landing spot.....with energy margin....so, it doesn't matter.
 
You do realize this thread is about automating that, right?
so...when you pick your spot....you pick one you can make, right? and you don't know how the winds will change on your way down....do you? The algorithm can do the same....no diff.

It can iterate using energy management to converge on a "makeable" site. It ain't hard....or rocket science.o_O
 
so...when you pick your spot....you pick one you can make, right? and you don't know how the winds will change on your way down....do you? The algorithm can do the same....no diff.

It can iterate using energy management to converge on a site.

And can converge on a very wrong answer, or not converge at all. You need to analyze this problem MUCH more carefully to make such a strong conclusion. Numerics has a way of kicking you in the balls when you try to iterate without understanding it.

I often DO have some idea what the winds below are doing, from observing things like flocks of birds.

And, since when does the appropriateness of a landing spot depend primarily on winds? I usually look for stuff like open fields. Winds often blow uphill to nasty looking terrain (not all of it at high altitude).
 
Back
Top