Droning on...

[Broomstick]

As an example, the Space Shuttle uses 5 onboard computers. 4 redundant computers and one independent computer that was created without the involvement of the first design team. Similar safety precautions can easily be observed for commercial aircraft.

Sure... the only obstacle being money. Historically, airlines have a bad habit of cost-cutting a little too much at times.

Computers are best for routine situations. Humans are best for the unforeseen and several types of emergencies. I'm not in a hurry to see that partnership end.

Nor am I at the moment, but it will happen when it's ready.

My concern is that someone will try to rush the technology into service before it is ready.

Computers are also able to react much faster than humans, and monitor far more information than humans. I can easily foresee scenarios where a computer would outreach a human pilot.

Oh, sure - no question. The thing is, for unforeseen occurrences for which the computer has no programming humans might be the better option. Hence, we still have them on board.

The only time an IFR (Instrument Flight Rules, as opposed to VFR, Visual Flight Rules) pilot really needs to be able to see is to land.

If the airport is equipped for CAT IIIc landings and the aircraft has a fail-operational autopilot (meaning that even if the autopilot experiences a failure below alert height it can still complete the landing without human assistance) then the pilots don't even need to see to do that.

True. So called zero-zero landings are already a reality. Such systems are still maintenance intensive, though. VERY useful, but they do cost money and require all parts of the system to function without flaw.

[Broomstick]

After such an incident, what the humans improvise can then be later studied and incorporated into flight knowledge, both for humans and machines. The two situations where that happened that I can recall offhand were the Airtransit flight that ran out of fuel over the Atlantic where the pilots on board were heard on the CVR referring to the successful unpowered landing in the Gimli Glider incident, and the the DHL cargo plane hit by a missile at the Baghdad airport that, if I recall correctly, utilized knowledge regarding controlling an airplane with differential thrust after hydraulic failure gathered from the Sioux City, Iowa accident. In the latter case, I believe some of that was even programed into the airplane systems.

I dunno about anything being programed in, its possible, but they knew they had to drop the landing gear to help slow down the plane while using a lot of engine thrust because of the Sioux City crash. Frankly they also just got damn lucky, that on top of it all, that the gear didn't collapse after they ran off the runway and they didn't strike any of the numerous UXOs all over the airport. It took something like a year to finish sweeping the place.

Well, yes, anytime part of your airplane is blown up there is an element of luck involved in getting down safely. Even so, it's an example of why pilots are trained to keep working on the problem and to never give up. Aircraft have landed after some pretty spectacular failures.

[Sea Skimmer]

Well, yes, anytime part of your airplane is blown up there is an element of luck involved in getting down safely. Even so, it's an example of why pilots are trained to keep working on the problem and to never give up. Aircraft have landed after some pretty spectacular failures.

Aye that they have, the RAF incident in which two trainers collided and then landed while still jammed together pretty well proved that one.

[LionElJonson]

As an example, the Space Shuttle uses 5 onboard computers. 4 redundant computers and one independent computer that was created without the involvement of the first design team. Similar safety precautions can easily be observed for commercial aircraft.

Sure... the only obstacle being money. Historically, airlines have a bad habit of cost-cutting a little too much at times.

The aviation industry (or, at least, the regulators thereof) love redundancy. If something is vitally important for the safety of the plane, odds are the planes got two or three of them. It's the safest way to travel for a reason, and that's not just because they analyze every accident for ways to make aviation safer. It's also because for every accident, there'll be a chain of four or five failures before things get to the point the plane goes down.

Computers are also able to react much faster than humans, and monitor far more information than humans. I can easily foresee scenarios where a computer would outreach a human pilot.

Oh, sure - no question. The thing is, for unforeseen occurrences for which the computer has no programming humans might be the better option. Hence, we still have them on board.

Yeah, and even if they could be piloted from the ground, the FAA and similar organizations will still probably require human pilots on board in case the radio fails and then the aircraft undergoes an emergency.

The only time an IFR (Instrument Flight Rules, as opposed to VFR, Visual Flight Rules) pilot really needs to be able to see is to land.

If the airport is equipped for CAT IIIc landings and the aircraft has a fail-operational autopilot (meaning that even if the autopilot experiences a failure below alert height it can still complete the landing without human assistance) then the pilots don't even need to see to do that.

True. So called zero-zero landings are already a reality. Such systems are still maintenance intensive, though. VERY useful, but they do cost money and require all parts of the system to function without flaw.

My understanding from my aviation professor (who's an ex-Air Force instructor) is that you're required by regulation to be able to visually see the runway before you attempt to land. That might just be an Australian rule, though.

[Broomstick]

As an example, the Space Shuttle uses 5 onboard computers. 4 redundant computers and one independent computer that was created without the involvement of the first design team. Similar safety precautions can easily be observed for commercial aircraft.

Sure... the only obstacle being money. Historically, airlines have a bad habit of cost-cutting a little too much at times.

The aviation industry (or, at least, the regulators thereof) love redundancy. If something is vitally important for the safety of the plane, odds are the planes got two or three of them. It's the safest way to travel for a reason, and that's not just because they analyze every accident for ways to make aviation safer. It's also because for every accident, there'll be a chain of four or five failures before things get to the point the plane goes down.

Well, yes, understood. Nonetheless, while the industry as a whole is in love with redundancy there have been some rather notable exceptions to that rule. The US had a rather spectacular DC-10 crash over Chicago a couple decades ago, for example, that was triggered in part by maintenance being pressured to perform their job quicker and cheaper, leading to an engine falling off an airplane and severing hydraulic lines and various other problems. And that's a major player, one of the operations that usually do things right and have the funds to pay for what's needed. Smaller operators... it's a problem. Aviation is expensive, the temptation to save money is always there. Even the wealthiest players tend to resist new regulations and new requirements on a basis of cost.

Computers are also able to react much faster than humans, and monitor far more information than humans. I can easily foresee scenarios where a computer would outreach a human pilot.

Oh, sure - no question. The thing is, for unforeseen occurrences for which the computer has no programming humans might be the better option. Hence, we still have them on board.

Yeah, and even if they could be piloted from the ground, the FAA and similar organizations will still probably require human pilots on board in case the radio fails and then the aircraft undergoes an emergency.

Until we can get on-board AI that functions well in emergencies, yes, that is probably the case.

True. So called zero-zero landings are already a reality. Such systems are still maintenance intensive, though. VERY useful, but they do cost money and require all parts of the system to function without flaw.

My understanding from my aviation professor (who's an ex-Air Force instructor) is that you're required by regulation to be able to visually see the runway before you attempt to land. That might just be an Australian rule, though.

My understanding is from speaking with US pilots who have been trained to do them. Right now, that's either military or some of the very high-end civilian air liners. Not only does the pilot(s) have to be trained and current, but the proper machinery has to be installed both on the airplane and on the ground, both of which require frequent maintenance. That means that while possible they aren't particularly common. For the most part, what you say is true - the pilot has to, at some point, be able to see the ground (what we aviation folks call "decision height").

On factor that hardware-lovers often fail to consider is that the airplane+runway is NOT a closed system. It's one think to run computer simulations and wank on about automated landings - quite another to deal with the real world where runways aren't always pristine. Could be anything from debris on the runway (what took down the Concorde, although that was a takeoff, not a landing) to flocks of birds to people on the runway when they aren't supposed to be there, whether that's a ground support vehicle that got lost or failed to properly follow directions or some crazy naked guy running around the airport (happened at Chicago Midway about three years ago. In February. Naked guy running around on the runways in knee-high snowdrifts. Yes, alcohol and drugs were involved). With the normal visual requirements of instruments landings such aberrations may well be spotted along with the runway. On a zero-zero landing you can't, by definition, see where you're going. Yes, radar helps considerably, but it's only been relatively recently we've had anything that starts to make scanning the runway ahead practical.

Any pilot at any level of aviation will have multiple stories of aborting landings due to something being the runway that shouldn't. Ask me about the go-karts in Wisconsin sometime.

Anyhow - I fly under US FAA regs due to an accident of geography - that's where I live. Australia, as you point out, has its own rules and it may be that such landings are not (yet) permitted there. It could be a reason as simple as your airports not having the hardware installed.

[Werrf]

and the the DHL cargo plane hit by a missile at the Baghdad airport that, if I recall correctly, utilized knowledge regarding controlling an airplane with differential thrust after hydraulic failure gathered from the Sioux City, Iowa accident.

While I do agree with you, this actually isn't a great example of human superiority over automated systems - I believe* that before the DHL incident, inspired by Sioux City and a similar incident in Japan, a system had been developed, tested and readied for installation that could take pilot input and translate it into control commands using engine thrust instead of ailerons/elevators, to allow pretty good control in the event of a loss of all hydraulic systems. The trouble is, it wasn't put in place on many aircraft because the situation was held to be too unlikely to be worth the expense of fitting it.

That said, we're not ready for full automation yet. I've spent far too long working with programming computers to think the kind of adaptive intelligence we'd need for a truly reliable autopilot is anywhere on the horizon.

*I know the system exists, I'm just not sure about when it was developed - I know the primary motivator was Sioux City and the Japanese incident.

[Broomstick]

and the the DHL cargo plane hit by a missile at the Baghdad airport that, if I recall correctly, utilized knowledge regarding controlling an airplane with differential thrust after hydraulic failure gathered from the Sioux City, Iowa accident.

While I do agree with you, this actually isn't a great example of human superiority over automated systems - I believe* that before the DHL incident, inspired by Sioux City and a similar incident in Japan, a system had been developed, tested and readied for installation that could take pilot input and translate it into control commands using engine thrust instead of ailerons/elevators, to allow pretty good control in the event of a loss of all hydraulic systems.

I think you missed the fine detail of my point - it was Sioux City, IA that showed that this could be done, and the black box recordings showed how. Although they were never able to duplicate the exact conditions of the accident (even with the crew who successfully survived the actual accident) the knowledge gained from that was incorporated into crew training AND fly-by-wire systems such as used by Airbus. Humans came up with the innovation while under duress, now that is being incorporated into systems. It was an important factor in the survival of the DHL airplane, although not the only one. Both the crew - who knew such techniques were possible - and the airplane's systems enabled a safe landing.

By the way - a picture of the missile damage. It wasn't trivial:

The trouble is, it wasn't put in place on many aircraft because the situation was held to be too unlikely to be worth the expense of fitting it.

Yes. Hence my point about people not being willing to pay for a truly capable flight control system for a drone aircraft. They're expensive!

I do wish to issue a correction - I erroneously thought the airplane in question was a B-747. It was actually an Airbus 300 cargo plane.

That said, we're not ready for full automation yet. I've spent far too long working with programming computers to think the kind of adaptive intelligence we'd need for a truly reliable autopilot is anywhere on the horizon.

The other part of the problem is that people who aren't in the know usually have an overinflated idea of how advanced our AI systems are, and underestimate how messy and troublesome the real world is.