How soon the pilotless airliner?
Join Date: Jun 2001
Location: UK
Posts: 37
Likes: 0
Received 0 Likes
on
0 Posts
I agree totally with all of what tdracer said in post #25, and in particular this bit:
'........but the day will come when letting a human decide is demonstrably less safe than letting the automatics decide. When we get to that point, it would be irresponsible to let humans drive with hundreds of people in the back.'
I'm from the avionics / computing side of things.
An example of how things evolve:
From 1970 onwards I was involved in the manufacture and testing of aviation syncros and actuators for fuel control systems on all the then - currrent military planes + Concorde (which had military-rated engines). These things cost a lot of money to make because they were lumps of metal with electrical windings, all of which had to be very accurately made to give the correct outputs. We were taking something that was intrinsically dumb and trying to calibrate it to give it the impression of intelligence by machining stuff to a few tenths of 1/1000th of an inch on a case by case basis. Go too far and it's scrap.
But these outputs were just there to give an indication of the position of a rotor relative to a starting point (how far did the pilot advance the throttles?), and sometimes how quickly that was being done (was he in a big hurry?). Very expensive to make because we're trying to get accurate info from a dumb device. And there's a chain of them, not just the one, and errors are additive.
Even 40 - odd years ago we knew that things would have to go from a 'lump of iron that's as intelligent as we can make it seem' to 'a lump of iron that only has to be able to revolve, but has sensors - as many as required for the application - but still vastly cheaper and more reliable, and then let a computer read the sensors and DECIDE what to do.
In those days, because we were involved in technology, we knew that this particular dream would eventually come about for two reasons: we knew that the then - nascent computing power was bound to get better, and also because it needed to, simply because we would get something much more accurate and cheaper at the same time. And even then, the computing power available could cope with that, but we had no devices available to put on the end of it. These days we do. But they aren't in planes.
The problem is that most companies can't go the 'how can we get the result we need, but in a totally different and better way' route. One reason for that is obvious; they are making money from how they do it now, and don't want anything to change. Also, there would have to be a large engineering change at the customers' end. Everyone is locked in. But also it's because most people don't think like that. Those who do become billionaires these days.
Clive Sinclair was a case in point of precisely how to do it wrong. He wanted to make a tiny tv. So he spent forever trying to figure out how to take a cathode ray tube (v. old tech) and bend it at right angles. He was trying to make the existing thing better, instead of inventing the LED screen.
Pilotless planes were a thing for the far future. But the future is nearer now, and it gets nearer at an exponential rate.
And that is the point. Computers are so powerful, and software has got so much better than we'd ever imagined it would by this time. Well, we didn't know what we imagined; just think back 20 years and see the difference. Then, a phone was a thing on a table somewhere, that you used to talk to people, and nothing else. Look at them now.
A pilot knows what he knows. Suitably intelligent control systems potentially know what all the best pilots ever knew, and therefore what to do.
A pilot is a single mind; these systems are / will be the products of the minds of large numbers of highly intelligent people (wish I was one).
No-one could jam the plane, they'd have to jam whatever it might be talking to. But if it's autonomous it has everything it needs built in. And a total electrical failure can be got round with multiple redundancy; when was the last time a plane had a total failure? And how would it be different if that happened to a pilotless plane?
30 years from now, well, I don't suppose I'll be around, but I'm fairly sure the tech available then would even surprise me.
To make the point, the software used on the shuttle Challenger is on record as being the most heavily - tested software ever written. It was written in the 60s, by people sitting in a room. In the next room there were people whose job it was to try to break it. They kept going around this loop until nothing the guys in the other room could do would break it. They reckon the code cost $10,000 a line to write. Sadly, when they checked the telemetry after the disaster, they discovered that the software had spotted the fuel leak and was trying to shut down the fuel flow, and thus allow the possibility of the astronauts escaping, but unfortunately the valves it was trying to send signals to already didn't exist.
So I don't think software will ever be the problem - if they spend enough getting it right - and if it performs as well as a human then there's no argument against it. The software will always know far more than a human about what a plane is doing, and far faster. It's the decision-making that has to be concentrated on. The whole thing about suddenly deciding to do something, and better than a human is, in principle, a question of hard work and testing the software to destruction time and again until it doesn't get destroyed. At that point it's at least as good as a human.
The ultimate goal is to have an autonomous system that can think and react in all situations like Sully did in one situation. In principle that's possible now.
'........but the day will come when letting a human decide is demonstrably less safe than letting the automatics decide. When we get to that point, it would be irresponsible to let humans drive with hundreds of people in the back.'
I'm from the avionics / computing side of things.
An example of how things evolve:
From 1970 onwards I was involved in the manufacture and testing of aviation syncros and actuators for fuel control systems on all the then - currrent military planes + Concorde (which had military-rated engines). These things cost a lot of money to make because they were lumps of metal with electrical windings, all of which had to be very accurately made to give the correct outputs. We were taking something that was intrinsically dumb and trying to calibrate it to give it the impression of intelligence by machining stuff to a few tenths of 1/1000th of an inch on a case by case basis. Go too far and it's scrap.
But these outputs were just there to give an indication of the position of a rotor relative to a starting point (how far did the pilot advance the throttles?), and sometimes how quickly that was being done (was he in a big hurry?). Very expensive to make because we're trying to get accurate info from a dumb device. And there's a chain of them, not just the one, and errors are additive.
Even 40 - odd years ago we knew that things would have to go from a 'lump of iron that's as intelligent as we can make it seem' to 'a lump of iron that only has to be able to revolve, but has sensors - as many as required for the application - but still vastly cheaper and more reliable, and then let a computer read the sensors and DECIDE what to do.
In those days, because we were involved in technology, we knew that this particular dream would eventually come about for two reasons: we knew that the then - nascent computing power was bound to get better, and also because it needed to, simply because we would get something much more accurate and cheaper at the same time. And even then, the computing power available could cope with that, but we had no devices available to put on the end of it. These days we do. But they aren't in planes.
The problem is that most companies can't go the 'how can we get the result we need, but in a totally different and better way' route. One reason for that is obvious; they are making money from how they do it now, and don't want anything to change. Also, there would have to be a large engineering change at the customers' end. Everyone is locked in. But also it's because most people don't think like that. Those who do become billionaires these days.
Clive Sinclair was a case in point of precisely how to do it wrong. He wanted to make a tiny tv. So he spent forever trying to figure out how to take a cathode ray tube (v. old tech) and bend it at right angles. He was trying to make the existing thing better, instead of inventing the LED screen.
Pilotless planes were a thing for the far future. But the future is nearer now, and it gets nearer at an exponential rate.
And that is the point. Computers are so powerful, and software has got so much better than we'd ever imagined it would by this time. Well, we didn't know what we imagined; just think back 20 years and see the difference. Then, a phone was a thing on a table somewhere, that you used to talk to people, and nothing else. Look at them now.
A pilot knows what he knows. Suitably intelligent control systems potentially know what all the best pilots ever knew, and therefore what to do.
A pilot is a single mind; these systems are / will be the products of the minds of large numbers of highly intelligent people (wish I was one).
No-one could jam the plane, they'd have to jam whatever it might be talking to. But if it's autonomous it has everything it needs built in. And a total electrical failure can be got round with multiple redundancy; when was the last time a plane had a total failure? And how would it be different if that happened to a pilotless plane?
30 years from now, well, I don't suppose I'll be around, but I'm fairly sure the tech available then would even surprise me.
To make the point, the software used on the shuttle Challenger is on record as being the most heavily - tested software ever written. It was written in the 60s, by people sitting in a room. In the next room there were people whose job it was to try to break it. They kept going around this loop until nothing the guys in the other room could do would break it. They reckon the code cost $10,000 a line to write. Sadly, when they checked the telemetry after the disaster, they discovered that the software had spotted the fuel leak and was trying to shut down the fuel flow, and thus allow the possibility of the astronauts escaping, but unfortunately the valves it was trying to send signals to already didn't exist.
So I don't think software will ever be the problem - if they spend enough getting it right - and if it performs as well as a human then there's no argument against it. The software will always know far more than a human about what a plane is doing, and far faster. It's the decision-making that has to be concentrated on. The whole thing about suddenly deciding to do something, and better than a human is, in principle, a question of hard work and testing the software to destruction time and again until it doesn't get destroyed. At that point it's at least as good as a human.
The ultimate goal is to have an autonomous system that can think and react in all situations like Sully did in one situation. In principle that's possible now.
Last edited by UPP; 22nd Aug 2013 at 20:55.
Before they start manufacturing pilotless airliners, I would love to see Airbus and Boeing manufacture a plane without serious design problems (787...cough...cough...).
Don't these pilotless aircraft require an airport with CAT III ILS when it comes to landing?
Don't these pilotless aircraft require an airport with CAT III ILS when it comes to landing?
Join Date: Jul 2013
Location: UK
Age: 68
Posts: 736
Likes: 0
Received 0 Likes
on
0 Posts
What if an airport service vehicle, waiting to cross a runway after a plane had landed, had a tech or driver problem (eg heart attack) and surged out onto the runway in front of the plane?
What if one plane lost control while landing or taking off and skidded towards another on a parallel runway, perhaps a bit like Asiana 214?
Could the system make the right choices here? The pilotless heli landing is a fine achievement but it did not have to share its required space with other traffic.
Even if one day we have a system which really can cope with every possible event, there is still the problem of selling it to the public. In London we have hundreds of planes flying over us all day, if the population was told "next year it will all be "pilot-less" there would be an uproar, and it would be fueled by the more sensationalist members of the press. It could prove difficult to get people on the ground to accept it, as well as the passengers.
In this world accidents will always happen, but imagine the outcry the day the first pilot-less airliner hits buildings.
What if one plane lost control while landing or taking off and skidded towards another on a parallel runway, perhaps a bit like Asiana 214?
Could the system make the right choices here? The pilotless heli landing is a fine achievement but it did not have to share its required space with other traffic.
Even if one day we have a system which really can cope with every possible event, there is still the problem of selling it to the public. In London we have hundreds of planes flying over us all day, if the population was told "next year it will all be "pilot-less" there would be an uproar, and it would be fueled by the more sensationalist members of the press. It could prove difficult to get people on the ground to accept it, as well as the passengers.
In this world accidents will always happen, but imagine the outcry the day the first pilot-less airliner hits buildings.
Join Date: Jun 2001
Location: UK
Posts: 37
Likes: 0
Received 0 Likes
on
0 Posts
'What if an airport service vehicle, waiting to cross a runway after a plane had landed, had a tech or driver problem (eg heart attack) and surged out onto the runway in front of the plane?'
Simple. A human pilot might see it, and might eventually react. Fast enough to avoid an accident? Who knows.
A computer-controlled plane would spot it and would attempt to react like greased lightning. Fast enough to avoid an accident? Who knows. But it would have a better chance than a human.
As an aside, we should actually use chimps to fly planes, when it comes to visual acuity. I saw a jaw-dropping demo once.
A chimp had a grid flashed up on a screen in front of him. Some of the squares in the grid were lit up. After a few seconds they were switched off. To get his reward, the chimp had to touch all the squares that had been lit up. He did. The interval got ever shorter until, eventually, human eyes never saw them light up in the first place. But the chimp always got it right.
A suitably-equipped plane will always be faster than you could ever be in trying to avoid this sort of accident.
Remember, sensors far better than a human can be made for anything you need to know. It's the software that matters.
Simple. A human pilot might see it, and might eventually react. Fast enough to avoid an accident? Who knows.
A computer-controlled plane would spot it and would attempt to react like greased lightning. Fast enough to avoid an accident? Who knows. But it would have a better chance than a human.
As an aside, we should actually use chimps to fly planes, when it comes to visual acuity. I saw a jaw-dropping demo once.
A chimp had a grid flashed up on a screen in front of him. Some of the squares in the grid were lit up. After a few seconds they were switched off. To get his reward, the chimp had to touch all the squares that had been lit up. He did. The interval got ever shorter until, eventually, human eyes never saw them light up in the first place. But the chimp always got it right.
A suitably-equipped plane will always be faster than you could ever be in trying to avoid this sort of accident.
Remember, sensors far better than a human can be made for anything you need to know. It's the software that matters.
Join Date: May 2008
Location: London
Posts: 4
Likes: 0
Received 0 Likes
on
0 Posts
Risk analysis
Soyuz / ISS rendezvous has now been fully automated to give a 4-orbit ascent and docking in under 6 hours. Ops efficiencies will drive implementation of automation in air flight too resulting in piecemeal introductions - easiest and greatest savings first for commercial contexts. A similar time-line will evolve for the military with differing concepts of what constitutes a saving. Once in that future the greatest threat to air safety becomes each still-human-operated rig which might decide to not co-operate with the global airborne separation system (GASS). Following risk analysis and agreements with unions crew in the pointy-bit will be retired from service and everyone in the air will feel much safer knowing they're not going to have a couple of human-beings being human at inappropriate moments (eg, caught up in a discussion about company policy and cruising on past the airport while GASS directs everything over, under and around them).
Not just commercial cargo. I can see this also being appealing for the military, who could deliver supplies to war zones without risking crew.
deltayankee: Already being done. 2 airframes have been operating in afghanistan for at least a year IIRC.
Lockheed Martin · U.S. Marine Corps to Keep K-Max Unmanned Cargo Re-Supply Helicopter in Theater for Second Deployment Extension
deltayankee: Already being done. 2 airframes have been operating in afghanistan for at least a year IIRC.
Lockheed Martin · U.S. Marine Corps to Keep K-Max Unmanned Cargo Re-Supply Helicopter in Theater for Second Deployment Extension
Nemo Me Impune Lacessit
Join Date: Jun 2004
Location: Derbyshire, England.
Posts: 4,095
Received 0 Likes
on
0 Posts
The technology doesn't even make it into the frame, never mind a poor fourth.
The first pilotless aircraft carrying a revenue load of pax will happen;
a). When all terrorism has been completely contained and
b). When insurance underwriters agree to fully cover passenger and third party legal liabilities for such an aircraft, (which currently run to billions of dollars of cover for piloted aircraft).
c). When a major airframe maker is prepared to invest billions of dollars in R & D for such a project believing they will have, at least, a break-even market (A380 still about 200 airframes away from break-even).
The actual cost saving from removing pilots will be miniscule when compared to the cost of R & D alone, never mind all the other associated costs like setting up ground control stations etc. etc.
This subject has been flogged to death so many times!
The first pilotless aircraft carrying a revenue load of pax will happen;
a). When all terrorism has been completely contained and
b). When insurance underwriters agree to fully cover passenger and third party legal liabilities for such an aircraft, (which currently run to billions of dollars of cover for piloted aircraft).
c). When a major airframe maker is prepared to invest billions of dollars in R & D for such a project believing they will have, at least, a break-even market (A380 still about 200 airframes away from break-even).
The actual cost saving from removing pilots will be miniscule when compared to the cost of R & D alone, never mind all the other associated costs like setting up ground control stations etc. etc.
This subject has been flogged to death so many times!
Last edited by parabellum; 22nd Aug 2013 at 23:12.
Join Date: Jul 2013
Location: redmond
Age: 88
Posts: 61
Likes: 0
Received 0 Likes
on
0 Posts
re challenger software
They reckon the code cost $10,000 a line to write. Sadly, when they checked the telemetry after the disaster, they discovered that the software had spotted the fuel leak and was trying to shut down the fuel flow, and thus allow the possibility of the astronauts escaping, but unfortunately the valves it was trying to send signals to already didn't exist.
Yes the lox hydrogen engines did have shutoff valves- but when the booster leak of hot gasses penetrated the center tank - it was already too late
Curious of source of yur comment however . . .
Join Date: Jun 2009
Location: Canada
Posts: 464
Likes: 0
Received 0 Likes
on
0 Posts
Docking with a space station is easy compared to landing an airliner, since you're moving a few feet or inches per second with no other traffic to get in the way, and has still been problematic to implement.
Join Date: Jun 2009
Location: Canada
Posts: 464
Likes: 0
Received 0 Likes
on
0 Posts
The Challenger transcript shows the engine management computers attempting a clean shutdown of the shuttle main engines during the breakup as they saw fuel pressure drop. I presume that's the source.
Edit: it's worth noting that we almost lost Challenger a few flights earlier due to defective engine sensors which caused the computer to shut down one engine during launch and wanted to shut down a second. The software appears to be easier than making sensors that work reliably, as we also saw in the AF447 case.
Edit: it's worth noting that we almost lost Challenger a few flights earlier due to defective engine sensors which caused the computer to shut down one engine during launch and wanted to shut down a second. The software appears to be easier than making sensors that work reliably, as we also saw in the AF447 case.
Last edited by MG23; 22nd Aug 2013 at 23:35.
Join Date: May 2008
Location: London
Posts: 4
Likes: 0
Received 0 Likes
on
0 Posts
Docking with a space station is like drawing alongside a jetty - you're moving slowly relative to it and you want to stop in a pre-determined place. Unlike parking next to a jetty you have to crash into a space-station which unlike a jetty is moving too. So it's important to manage the energy carried by the docking craft as there's no place for any excess to go except into changing the trajectory and/or attitude of the combined crafts. The automated docking sequence has no concept of self-preservation only parametric boundaries linked to sensor inputs plus any optional input from the PAX turning into crew. The ascent to the ISS orbit and chase is more equivalent to the landing phase for an aircraft - find the destination, alter altitude to reach it, manage energy and trajectory while not running out of fuel, making a bad decision or flipping the wrong switch. Prior to automation this phase took 2 days, now its 6 hours. In an article introducing the new regime, SpaceflightNow observed: "The fast-track rendezvous reduces the workforce required to operate the Soyuz spacecraft. Instead of staffing the Soyuz control center in Russia for more than two days, a full complement Soyuz controllers will only be needed for a day". While taxpayers are hardly likely to be rejoicing with the news of this saving, it does mean that the productivity of the industrial process of spaceflight is increased - more flights from the same cost of ground facilities and staff, from the same level of investment. Of course that is another way of interpreting the savings made by automation with respect to flight crew - the efficient deployment of capital within businesses flying freight or people or productivity gains for tax-payers in government run outfits.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Join Date: Jul 2013
Location: UK
Age: 68
Posts: 736
Likes: 0
Received 0 Likes
on
0 Posts
How about a scenario in which a mis-calculation leads an airliner to run out of fuel. The flight crew spot a disused airfield and manage to glide in safely, without harming any one on board or any of the public at a meeting on the runway. A disused airfield would not have the equipment for a pilot-less landing.
Or a scenario where an engine failure takes out the hydraulic lines and all flight controls. The flight crew gradually realise that they can keep partial control of direction and altitude by altering the thrust of one engine. They manage to make it to an airport and despite a crash landing there are significant numbers of survivors.
Of course, I am aware that sometimes flight crews have made th wrong decisions and caused an avoidable accident. All humans are fallible.
In my opinion human fallibility extends to those who design automatic systems and have total faith in the technology and total faith that they have taken all eventualities into account.
If/when a pilot-less plane does crash, and regardless of what the aviation industry thinks, the public outcry and insurance questions could be insurmountable.
Or a scenario where an engine failure takes out the hydraulic lines and all flight controls. The flight crew gradually realise that they can keep partial control of direction and altitude by altering the thrust of one engine. They manage to make it to an airport and despite a crash landing there are significant numbers of survivors.
Of course, I am aware that sometimes flight crews have made th wrong decisions and caused an avoidable accident. All humans are fallible.
In my opinion human fallibility extends to those who design automatic systems and have total faith in the technology and total faith that they have taken all eventualities into account.
If/when a pilot-less plane does crash, and regardless of what the aviation industry thinks, the public outcry and insurance questions could be insurmountable.
Join Date: Jun 2009
Location: Scotand
Age: 68
Posts: 57
Likes: 0
Received 0 Likes
on
0 Posts
Learmount link
In the link in post #1, Learmount says that the remote pilot will have access to data "in real time" Has he ever made a satellite phone call? The latency is about one second and this makes even phone calls very difficult for the uninitiated. Trying to remotely fly a plane with a one second delay in inputs and feedback will be almost impossible. The plane has to be autonomous in regards to flight surfaces etc. Only the routing and strategy could be controlled remotely.
Join Date: Jan 2001
Location: Home
Posts: 3,399
Likes: 0
Received 0 Likes
on
0 Posts
Joyride
If you watch the videos in my previous post you will see that autonomous vehicles are more than capable of landing at a disused airfield without any landing aids, in fact they have many advantages over a piloted aircraft in that situation, ie they can do it IMC/in the dark and unlike a pilot who has never practised for real can be programmed with the full glide performance etc
re your point about hydraulic failures etc.
There have I believe been three big occurrences. Sioux City, DHL Iraq and Korea?
The first went not too bad, the second went very well, and third was a total disaster.
Propulsion Controlled Aircraft (PCA) | NASA
The work has already been done to solve the problem for computers.
I reiterate.
The problems are in no way technical.
If you watch the videos in my previous post you will see that autonomous vehicles are more than capable of landing at a disused airfield without any landing aids, in fact they have many advantages over a piloted aircraft in that situation, ie they can do it IMC/in the dark and unlike a pilot who has never practised for real can be programmed with the full glide performance etc
re your point about hydraulic failures etc.
There have I believe been three big occurrences. Sioux City, DHL Iraq and Korea?
The first went not too bad, the second went very well, and third was a total disaster.
Propulsion Controlled Aircraft (PCA) | NASA
The work has already been done to solve the problem for computers.
I reiterate.
The problems are in no way technical.
Reference the Jetstream flight above here is a quote from a previous thread:-
Full thread here:-
Testflight with 'unmanned' Jetstream
Maybe not so "automated" as it's trumped up to be?
They are not testing auto take-off or auto-land - the two pilots on board will handle those phases of the flight
Testflight with 'unmanned' Jetstream
Maybe not so "automated" as it's trumped up to be?
Join Date: Apr 2008
Location: HK
Posts: 513
Likes: 0
Received 0 Likes
on
0 Posts
If you watch the videos in my previous post you will see that autonomous vehicles are more than capable of landing at a disused airfield without any landing aids, in fact they have many advantages over a piloted aircraft in that situation, ie they can do it IMC/in the dark and unlike a pilot who has never practised for real can be programmed with the full glide performance etc