Thought I'd share an article from this month's FlightCom
Known knows
Known unknowns
Unknown unkowns
We have ourselves a Rumsfeld fan.
-
Join the MyBroadband community
South Africa’s biggest forum. Discuss, discover, and connect with thousands of members.
-
Question of the Day: Will you be subscribing to the new Disney+ R49 streaming package?
Boeing looking into passenger planes without pilots
- Thread starter Kevin Lancaster
- Start date
-
- Tags
- boeing
Known knows
Known unknowns
Unknown unkowns
We have ourselves a Rumsfeld fan.
Indeed. Case in point:
A very good friend of mine was doing a sector between Melbourne and Hong Kong going northbound on the airway. They were gently cruising along when he heard another aircraft checking in on the same airway going the opposite direction. Upon hearing the aircraft's callsign and altitude he immediately checked the spot wind and requested a parallel offset from the airway. ATC approved the offset. They were barely properly offset when the Qantas A380 passed 1000ft overhead. They caught only the residual of the A380's wake but he described it as being hit by an invisible truck. Now imagine they caught the full brunt of that whale's wake turbulence. Would a computer know to do this? Would a computer have that type of situational awareness? Long answer, no.
Idiosyncratic
Expert Member
Haha, RVSM isn't always the coolest thing that has been implemented in aviation
There are pros and cons - as with anything.
In terms of the AI aspect, in theory one could teach AI (particularly now with ADS-B becoming more and more important) to offset for wake turb, but we're dealing with computer chips from inkjet printers in many of these machines (yes, really) and it's a long and tedious process to certify new chips.
What a lot of folks forget about aviation is regulatory red tape holding technology back. Back when I did initial training I did it on aircraft almost as old as what my grandfather was at the time. Not much as changed since then in that industry.
I think about a go-around I did on a flight recently. ATC cleared an aircraft to line up and take-off while the bird I was operating was on short final. I could see it coming...was like I could smell it was going to happen. No harm done though, he made a mistake and we responded accordingly. ATC allowed us to follow an unorthodox procedure to join on short final again within a minute or two. If AI was in control, assuming it went around, you'd be looking at a scenario where precious minutes would have been wasted following "rules", not to mention the correct procedure would have taken us onto a potential collision course with the aerie taking off. The total funds wasted just by such instances would pay pilot salaries the world over - several times.
How do you program AI to cut corners (breaking rules, laws and taking shortcuts) where necessary by preforming manoeuvres never done before? Assuming we achieve that, how are our regulatory bodies going to respond when they find out we're programming aircraft to intentionally break rules? They're going to have a field day. To a hammer, every problem seems like a nail.
Haha, RVSM isn't always the coolest thing that has been implemented in aviation
There are pros and cons - as with anything.
In terms of the AI aspect, in theory one could teach AI (particularly now with ADS-B becoming more and more important) to offset for wake turb, but we're dealing with computer chips from inkjet printers in many of these machines (yes, really) and it's a long and tedious process to certify new chips.
What a lot of folks forget about aviation is regulatory red tape holding technology back. Back when I did initial training I did it on aircraft almost as old as what my grandfather was at the time. Not much as changed since then in that industry.
I think about a go-around I did on a flight recently. ATC cleared an aircraft to line up and take-off while the bird I was operating was on short final. I could see it coming...was like I could smell it was going to happen. No harm done though, he made a mistake and we responded accordingly. ATC allowed us to follow an unorthodox procedure to join on short final again within a minute or two. If AI was in control, assuming it went around, you'd be looking at a scenario where precious minutes would have been wasted following "rules", not to mention the correct procedure would have taken us onto a potential collision course with the aerie taking off. The total funds wasted just by such instances would pay pilot salaries the world over - several times.
How do you program AI to cut corners (breaking rules, laws and taking shortcuts) where necessary by preforming manoeuvres never done before? Assuming we achieve that, how are our regulatory bodies going to respond when they find out we're programming aircraft to intentionally break rules? They're going to have a field day. To a hammer, every problem seems like a nail.
Pretty big mistake.
Idiosyncratic
Expert Member
It happens more often than one thinks. It's not a big deal if you have a pilot or two in the cockpit
The_Assimilator
Executive Member
It happens more often than one thinks. It's not a big deal if you have a pilot or two in the cockpit
And if everything was AI-controlled, there wouldn't have been that mistake to begin with.
The Voice
Honorary Master
I guess there is already a lot of automation involved when the plane's cruising - the pilot's main function is to take off and land in a perfect scenario.
Speaking of landing: had to catch a connecting flight to SA recently via Zurich. Flight was delayed, so the pilot put foot. Anyway the airport was snowed over so we circled a bit so that the ground crew could clear the runway. Before we touched down we hit some serious crosswinds - could actually feel the plane swaying from side to side. The pilot managed to land safely and everyone did that weird thing where they start clapping. All this got me wondering how an AI would react to that situation.
Speaking of landing: had to catch a connecting flight to SA recently via Zurich. Flight was delayed, so the pilot put foot. Anyway the airport was snowed over so we circled a bit so that the ground crew could clear the runway. Before we touched down we hit some serious crosswinds - could actually feel the plane swaying from side to side. The pilot managed to land safely and everyone did that weird thing where they start clapping. All this got me wondering how an AI would react to that situation.
mercurial
MyBB Legend
I don't think we can completely remove pilots. I think in those scenarios, there would need to be a manual takeover by the pilot on duty.
Idiosyncratic
Expert Member
I thought someone would bring that up, but that's where you're very very wrong. You're completely ignoring other facts which was brought to the table and cherry-picking things to respond to.
Airplanes don't have the ability to come to a complete stop to maintain 5nm (or in this case, 2 min) separation - a computer can't just say "hey, we're getting too close, let's pull over here for a minute". So you have several problems:
1. For our lifetimes at least, there'll be pilot controlled aircraft. You and I will be dead before not a single aircraft flying needs a pilot. ATC will be around for a while to come also. Mistakes happen. Having a hybrid of autonomous aircraft + ATC and human operated aircraft + ATC is asking for more trouble than we have already.
2. Computers screw up. All the time. Drones are a good example. This is now disregarding the fact that the flight computers responsible for flying aren't much faster (in terms of clock speed) than calculators.
3. Obstacles on the runway (not just aircraft, I've had to go around for all sorts) are a reality. Even in an AI controlled world, you'll still have cases where runways will be obstructed. Unless everything on the planet is on a single system (everything linked), you'll be sitting with an interesting situation. That's not to mention unreported potholes, or any FOD. This issue is also not just limited to the third world - but it's much much worse in Africa than elsewhere.
4. AI doesn't have the same abstract view or intuition that a human has. This is a critical element of flying.
5. I've never been on a completely uneventful flight. When things go wrong - and boy, they do all the time, how does the computer determine what is actually wrong? Once again, the Qantas A380 incident comes to mind. It was very manageable by the crew, but the computer would have killed everyone on board.
6. Maintenance mistakes can be fatal. They happen, and most of the time, pilots compensate for them. Forgot to reconnect a critical wire somewhere in the world of AI controlled airliners?
NarrowBandFtw
Honorary Master
Nitpicky, but funny: you know the plane's throttle is controlled by hand right?
I would imagine the AI overlord will determine if the hand clapping was to its satisfaction first and if so, open the doors, if not, take off and ram into the nearest mountain ... :twisted:
I wonder how AI would handle this little incident? Damn dugongs, always upsetting everything around them.
https://avherald.com/h?article=4a5e80f3
The_Assimilator
Executive Member
The AI responsible for plotting flight paths would know about the wake turbulence of an A380 and thus never schedule a path bringing another plane close to that turbulence.
Assuming the flight paths did somehow get scheduled in this way, the AI of the A380 and the AI of the business jet would communicate when closing, determine there would be an overlap, and automatically obtain clearance to alter their paths to prevent the incident.
Assuming the incident did actually occur despite both the above contingencies, the business jet AI would already have this prior incident recorded and would be able to use that information to recover from this one.
Assuming the business jet AI did not have this incident recorded, an AI would be able to react almost instantaneously to abnormal sensor readings and begin applying control adjustments to recover. An AI would also be able to apply far more granular control corrections, and far faster, than any human ever could. As a result, the AI pilot would be able to start recovering its plane from the wake turbulence even while still being subjected to it.
kat06sa
Expert Member
Guys please listen to Idiosyncratic. Haven't been back to mybb for a while. But from just reading the last 2 pages, Idiosyncratic knows what he/she is talking about. I am no authority in the industry, I just work there. And as just an ordinary worker bee. Pilot-less passenger planes are decades, maybe a century away. There are just too many factors that cannot be programmed.
As pilots, we understand you will judge our entire career on the last bad landing we had. But it's the experience that we've had and shared experience, something you never see, that will probably keep you safe, and keep the industry the safest transportation device in the world.
Although, our role may evolve. There is no way I'll put myself or a loved one on a pilot less plane.
As pilots, we understand you will judge our entire career on the last bad landing we had. But it's the experience that we've had and shared experience, something you never see, that will probably keep you safe, and keep the industry the safest transportation device in the world.
Although, our role may evolve. There is no way I'll put myself or a loved one on a pilot less plane.
kat06sa
Expert Member
This has to be a learned experience. Many aircraft have passed under the A380 without such drastic incidence.
Idiosyncratic
Expert Member
You realize what actually caused this incident to be so severe was AI?
What about CAT at that level? That's not always predictable either.
Thousands of flights have flown 1000 ft under an A380 without experiencing what they experienced during the incident.
And the above-mentioned technology (where the planes communicate) have existed for years - it's called ACAS (or TCAS in some countries). Using it in this case would have constituted "abuse of the system" and been a violation in itself. You're underestimating the stupidity of regulatory bodies.
^This part is a real jewel, because AI did react almost instantaneously. AI was flying the aircraft at the time of the incident. And AI knows how to handle turb, yet it helped cause several passengers to sustain pretty terrible injuries. After it realized it royalty screwed up, it dumped control to the pilots. What if it couldn't? You would have an aircraft that is doing what it thinks is logical to recover, when in actual fact it's making things much worse. Train the aircraft to recover correctly during wake turb, I hear you say? Well, thing is, the upset could have been caused by something else which would have been sensed as exactly the same by the aircraft - but with different effects and recovery techniques required.
Idiosyncratic
Expert Member
It's like you seem to be assuming three things with that statement:
1. Humans can't do that.
2. AI gets it right every time.
3. Control would be centralized. Where's the autonomy in that?