In Atmosphere Computer Navigation

[Spaced Out]

Why is it so hard to use computers for in atmospheric flight?

[Gaarst]

It isn't. Or do you mean navigation as in flight path? We have GPS and a whole bunch of navigation systems then.

Edited August 22, 2017 by Gaarst

[Exploro]

GPS-based navigation systems as well as inertial navigation systems are common place. These are further augmented by the various legacy ground-based navaids like VOR's. Thus aerial navigation systems provide many means by which to navigate an aircraft. It all a matter of learning how to use them and knowing their limitations for a successful flight.

Edited August 22, 2017 by Exploro

[YNM]

It isn't at all.

Altough, FMS isn't a flight computer in the sense that there's a flight computer in Soyuz that does all the launch, rendezvous, docking, undocking, and reentry. FMS is more for management from ground and as a reference for the crews. It's true that they do hold some data for automated navigation but some can be inserted manually into the actual automatization.

Surely you got autopilots nowadays. The only thing they haven't done automatizing is taxiing on ground. All the others can be carried out by the autopilot as long as you entered the right input (so TO/GA on takeoff I think) and have the right instrument infrastructure (so either RNAV w/ GPS or VOR/DME for SID/STAR departure and approach - this one have to use FMS apparently ; airtracks for over-ocean operations - also need FMS ; then ILS cat. II and higher for landing).

If you mean with "hard" as in "looks complicated", consider that there are much more flights than there are satellites. It's hard because it's quite crowded.

Edited August 22, 2017 by YNM

[Steel]

2 hours ago, Spaced Out said:

Why is it so hard to use computers for in atmospheric flight?

Are you talking planes or rockets?

[Nibb31]

Why do you think it is?

[p1t1o]

Whether you call it "hard" or "easy", a fly-by-wire control system need:

Access to multiple air data sensors for speed, angle of attack, dynamic and static pressures amongst others.

Access to accurate data on the aircrafts attitude (wary, lazy etc.   LOLJK)

A fast computer to make all of the calculations in real-time.

A large base of data generated by flight testing the airframe in all reasonably expected envelopes of flight.

 

So compared to designing a nice, stable single-engined prop monoplane, yeah, building an FBW aircraft is a lot harder. Much easier to fly though.

Its "easy" because of all of the work done in the mid-late 20th century and because we have the appropriate technology.

 

Hopefully it is obvious why a fly-by-wire system operating in vacuum and freefall is "easier".

[NSEP]

Because atmospheric flight has unpredictable weather and space has no significant unpredictable path changing things to worry about?

Quick geuss tho.

[kerbiloid]

Because human crowd is conservative and gets scared flying in a plane without an organic pilot.

Every phase of the flight is already automated, but the onlookers need somebody to rotate the handle.

Spoiler

They think, this is more safe.

Edited August 23, 2017 by kerbiloid

[p1t1o]

3 hours ago, kerbiloid said:

Because human crowd is conservative and gets scared flying in a plane without an organic pilot.

Every phase of the flight is already automated, but the onlookers need somebody to rotate the handle.

  Reveal hidden contents

They think, this is more safe.

Right. Air accidents are rare, but the overwhelming proportion of them are caused by human error.

But even knowing that, I'm still happier with a man-in-the-loop, go figure.

[Snark]

On 8/23/2017 at 7:26 AM, p1t1o said:

But even knowing that, I'm still happier with a man-in-the-loop, go figure.

Computers are much better than humans at handling routine tasks, repetitively, which have a standard set of best practices that are safe-but-boring, such that deviating from those best practices causes danger.

Humans are much better than computers at handling the unexpected.

A computer would be better than a human pilot at maintaining constant vigilance in the air, constantly monitoring collision warnings and transponder signals and all the airplane's various "health monitors" of its systems, and instantly sounding an alert when something's wrong.  A computer would be better at following a pre-programmed flight path.  And lots of other things.

But... sometimes the unexpected happens.  Go watch the movie Sully, at try to tell me that a computer could handle any of that:

• The humans had to make a judgment call:  land the plane on water, or try to make it to the airport?
• The humans understood what happened, i.e. that the engines failed because of a bird strike, not something internal to the plane that might affect how to respond.

...and humans are good at noticing "Something smells like a burning wire" or "What's that funny sound?" or "Passengers are reporting a loud bang" or "Hey, smoke appears to be coming out of the wing" or various such imminently-need-to-be-dealt-with problems that might not have an automated sensor that catches it.

So, I'm all for letting the machines mostly fly the planes... but really want there to be a highly-trained human right there with a hand near the controls.

[Green Baron]

Yeah, right. Also aeronautical navigation systems aren't stand-alone systems like a simple 2-axis autopilot. They are integrated in communication between aircrafts (like TCAS, Transponders, ...) and between aircraft and ground (like navigation aids, automated landings). Crew that handles these systems need special training more extensive than mere flying before they are allowed to push the respective buttons.

If they fail to respect the procedures things can go terribly wrong, despite of computers and traned crews. I recall the case of two airplanes crashing mid flight because one relied on TCAS and the other on the (distracted by other tasks) controller. Very sad that was.

Report

Edited August 24, 2017 by Green Baron

[wumpus]

1 hour ago, Snark said:

Computers are much better than humans at handling routine tasks, repetitively, which have a standard set of best practices that are safe-but-boring, such that deviating from those best practices causes danger.

The famous counter argument would be Flight 401, a plane that crashed because the entire flight crew was trying to diagnose a burned out light bulb.
http://articles.sun-sentinel.com/1992-12-29/news/9203090325_1_landing-gear-jumbo-jet-plane

Not to mention the presumed deliberate crash of Germanwings Flight 9525 in 2015.  MH370 (Malaysian plane that completely disappeared over the Indian Ocean) was worth several billion dollars in parts.  The exact reason it disappeared is likely to never be known, but piracy is certainly a possibility (it worked for BJ Cooper).  Presumably it would be easier to subvert a programmer than a pilot/flight crew (since it would be easier to kill the pilot when he landed).

I'm guessing that terrorist attacks/hacking are a better justification.  To commandeer a plane with a pilot, a terrorist has to force his way into the cockpit (and all passengers can fight him off).  To commandeer a pilotless aircraft, a hacker needs to find an exploit in the control software and fly the plane into the ground/buildings/whatever.  To a certain extent, this is already true of Airbus planes, and the also attempt to avoid crashing due to pilot error.  Early data shows that the Boeing system (trust the pilots) is better, but I suspect in the long run the Airbus software will remove bugs from their software (human error will always be with us). 

Note that the FAA (and presumably similar agencies worldwide) actually has standards for software.  You can't put a windows machine in charge of your aircraft: *every* line of code needs a justification for its existence (don't expect this to be true for automobile "autopilots").  Terrorist hackers are much more likely to find easier targets elsewhere (such as automobile/truck "autopilots").

One final thing: if you want to argue humans would do better in "Sully" like situations, it might be better to hand over the controls to a "harbor pilot" on the ground [and use encrypted radio controls] with "impossible" amounts of takeoff/landing experience (for that airport) relative to normal pilots (much the way that harbor pilots have the advantage over normal captains in their harbor).  Assuming they would be physically located at "their" airports, latency wouldn't be an issue.

[Snark]

32 minutes ago, wumpus said:

One final thing: if you want to argue humans would do better in "Sully" like situations, it might be better to hand over the controls to a "harbor pilot" on the ground [and use encrypted radio controls] with "impossible" amounts of takeoff/landing experience (for that airport) relative to normal pilots (much the way that harbor pilots have the advantage over normal captains in their harbor).  Assuming they would be physically located at "their" airports, latency wouldn't be an issue.

Maybe... but sure would suck if someone jammed the signal.  Or there was a power outage on the ground.  Or if whatever-catastrophe-it-was on the plane fried the radio.

There's also one final consideration:  When you ride on a plane, you're trusting your life to a machine with zillions of moving parts, that has to be maintained and serviced by humans.  You have to just trust those humans.

The person I trust the most, to stop <bad thing> from happening, is the person who personally has the most to lose if <bad thing> happens.  I'm sure people on the ground are professional and dedicated, but if something goes wrong on the plane, it's not their personal skin on the line.  A pilot is trusting his own life, along with the passengers'.  No matter how much your innate nobleness of character; no matter how diligent your professional dedication; no matter what else, I would imagine that "...and I die if this breaks" must really help to ensure and maintain focus.

[p1t1o]

14 hours ago, Snark said:

Computers are much better than humans at handling routine tasks, repetitively, which have a standard set of best practices that are safe-but-boring, such that deviating from those best practices causes danger.

Humans are much better than computers at handling the unexpected.

A computer would be better than a human pilot at maintaining constant vigilance in the air, constantly monitoring collision warnings and transponder signals and all the airplane's various "health monitors" of its systems, and instantly sounding an alert when something's wrong.  A computer would be better at following a pre-programmed flight path.  And lots of other things.

But... sometimes the unexpected happens.  Go watch the movie Sully, at try to tell me that a computer could handle any of that:

• The humans had to make a judgment call:  land the plane on water, or try to make it to the airport?
• The humans understood what happened, i.e. that the engines failed because of a bird strike, not something internal to the plane that might affect how to respond.

...and humans are good at noticing "Something smells like a burning wire" or "What's that funny sound?" or "Passengers are reporting a loud bang" or "Hey, smoke appears to be coming out of the wing" or various such imminently-need-to-be-dealt-with problems that might not have an automated sensor that catches it.

So, I'm all for letting the machines mostly fly the planes... but really want there to be a highly-trained human right there with a hand near the controls.

Excellent way of putting it. 

[YNM]

13 hours ago, wumpus said:

The famous counter argument would be Flight 401, a plane that crashed because the entire flight crew was trying to diagnose a burned out light bulb.
http://articles.sun-sentinel.com/1992-12-29/news/9203090325_1_landing-gear-jumbo-jet-plane

That's the crew problem. If they were to be scrapped altogether the system would continue running. The crash were purely because of the crew's mistakes.

More examples :

https://en.m.wikipedia.org/wiki/TAM_Airlines_Flight_3054 - crash due to incorrect setting of the autopilot.

https://en.m.wikipedia.org/wiki/Philippine_Airlines_Flight_137 - another one.

There are a lot more of crashes caused by improper usage or communication. The amount that's caused by the automatic system itself without any intervention is much lower, even closer to non-existant.

[Exploro]

On 8/24/2017 at 1:20 PM, Snark said:

Humans are much better than computers at handling the unexpected.

Provided humans recognize an unexpected condition is occurring, can correctly identify what exactly is happening and take the appropriate corrective actions in a timely way.

Air France Flight 447 is an example where these did not happen.

Edited August 25, 2017 by Exploro

[Snark]

13 minutes ago, Exploro said:

Provided humans recognize an unexpected condition is occurring, can correctly identify what exactly is happening and take the appropriate corrective actions in a timely way.

Air France Flight 447 is an example where these did not happen.

Yep.  But there are other occasions when they do help.  Which is why I want both a smart, vigilant computer and a highly-trained human sitting in the cockpit.

[razark]

10 minutes ago, Exploro said:

Provided humans recognize an unexpected condition is occurring, can correctly identify what exactly is happening and take the appropriate corrective actions in a timely way.

Isn't that exactly what a computer would have to do?

Except the computer can't think outside the box like a human can.

[Exploro]

10 minutes ago, razark said:

Isn't that exactly what a computer would have to do?

Except the computer can't think outside the box like a human can.

In the case of 447 the computer did recognize there was a problem. When icing blocked the pitot tubes there was a fault in the air data the flight computer was receiving from these sources. By design the flight computer switched off the autopilot and converted flight controls from a normal flight mode to an alternate flight control mode. The computer was designed to hand over control back to the flight crew; which it did. The switch off assumes the crew would do the things I described in my last post. However we know the crew failed to impliment the procedures necessary to cope with erroneous air speed data and executed the wrong manuevers that caused a stall and kept the plane in a stalled condition that resulted in the crash.

Any way, computers may not be able to think outside the box...but they are immune to the physiological and congnitive pitfalls that plauge people.

Edited August 26, 2017 by Exploro

[YNM]

I have to add, though : even if it's true that computers are "better" at the job overall, sometimes weird human notices gathered from million(s) of years of living ancestors gives us weird insights not really applicable with computers.

Although it's a bit like chicken-egg problem : if humans haven't built them in, then yes the machines can't notice it.

Edited August 26, 2017 by YNM

[razark]

13 minutes ago, Exploro said:

The computer was designed to hand over control back to the flight crew...

And what would the computer have done?

It shut off control exactly because it did not have valid data, and could not operate the aircraft.  Granted, in this case, the crew didn't do the right thing, but a different crew may have been able to do something else.

 

What would the computer have done in the absence of valid data on which to operate, and no way to get that data because it's neither programmed to nor equipped with the sensors needed?

In this case, the computer did "recognize an unexpected condition is occurring", but could a computer have "correctly identif[ied] what exactly is happening and take the appropriate corrective actions in a timely way"?

Edited August 26, 2017 by razark

[Exploro]

28 minutes ago, razark said:

And what would the computer have done?

It shut off control exactly because it did not have valid data, and could not operate the aircraft.  Granted, in this case, the crew didn't do the right thing, but a different crew may have been able to do something else.

 

What would the computer have done in the absence of valid data on which to operate, and no way to get that data because it's neither programmed to nor equipped with the sensors needed?

Did I not just say what the computer would have done; or rather did in fact do? It transferred control over to the flight crew...however it does so without any way of knowing whether or not the flight crew is prepared to takeover. You might be correct that a different flight crew might have done something differently and recovered the situation. But the fact of the matter is; and it is the thing that matters the most, the crew of Flight 447 did not respond correctly to an upsetting situation....supposedly the thing we are good at.

As the flight control system is designed to automatically revert to a manual control mode it makes speculating on how the computer would have handled it moot. It was not designed to deal with such situations to begin with. If one were I'd say the thing it would do was ignore air speed data and use for instance angle of attack data. Like any airfoil; an airplane's wing will stall at a specific critical angle of attack. This critical AoA does not change with airspeed. Thus a computer can use AoA data to keep the plane from exceeding critical AoA and thus prevent stall in the absence of reliable air speed information.

Edited August 26, 2017 by Exploro

[razark]

1 minute ago, Exploro said:

Did I not just say what the computer would have done; or rather did in fact do? It transferred control over to the flight crew...

Excuse me, I was unclear.  The question I meant to ask was "What should the computer flying the aircraft do, when it has invalid data, and there is no one to give control to?"

 

4 minutes ago, Exploro said:

...the crew of Flight 447 did not respond correctly to an upsetting situation...

The problem was what the crew did (or didn't) do, right.  Remove them from the loop and assume an automated flight for the purposes of the exercise.

 

11 minutes ago, Exploro said:

If one were I'd say the thing it would do was ignore air speed data and use for instance angle of attack data.

We've already established that some of the data the computer is receiving is invalid.  Should the computer trust the AoA data?  Altitude?  Attitude?  How should the computer determine what is instrument/sensor failure vs. correct data?

What about other situations?  If the aircraft loses engine power, how should a computer react?  How does it decide on the best place to attempt a landing?

 

There's a lot of questions involved before turning automated flight over to a computer that can be handled by a (well trained) human flight crew.

[Exploro]

39 minutes ago, razark said:

We've already established that some of the data the computer is receiving is invalid.  Should the computer trust the AoA data?  Altitude?  Attitude?  How should the computer determine what is instrument/sensor failure vs. correct data?

In the case of 447 only one source of data was erroneous; the air speed data.

Among the things computers are very good is fault checking. The conditions prior to an upset would be a known and thus makes comparing what all other sensors read before the upset occurred and what they are reading now a simple task. If air speed data became unreliable; the computer could compare among other things the AoA before the upset and what it is reading now. If they are the same than the computer could deduce the reading are valid and can be trusted. It can also compare them to other sources too. For instance if an airplane is in level cruise flight and suddenly airspeed shows an increase; yet power setting was not increased nor any command to pitch down sent to the flight surfaces, and AoA is unchanged, the Computer once again knows air speed data is wrong and can be ignored.

Edited August 26, 2017 by Exploro