Is automated flight too uncertain, too complex, and too poorly supported?

[Phil Bunch]

After studying the very interesting threads in this forum regarding some fine points in Boeing's flight management systems, I have some questions that have been increasingly bothering me.  From a technical standpoint, and as a 747-400 enthusiast, I really enjoy reading the discussions among the many experts who participate in PSX, of course.

This thread is an example:

http://aerowinx.com/forum/topic.php?id=2207

First, given the presence of uncertainties with respect to how an airliner will respond to various combinations of settings and parameters, how does anyone comfortably and with adequate certainty fly an airliner!?

These discussions have made me less comfortable as an occasional passenger - the last things I would want as a passenger are (1) pilots with unnecessary (?) complexity, uncertainty, and ambiguity to cope with, (2) lack of access to aircraft software updates due to cost, and (3) failure paths that are preventable.  Am I overreacting by occasionally being distressed by these most interesting technical discussions?  Would these hypothetical improvements make any difference to aviation safety?

However, there are so few accidents in the airline industry that it is all but impossible to experience injury or death due to an airliner crashing.   We are truly blessed with an amazingly safe transportation system, vastly safer than for example traveling by automobile.  This point of view argues that our current practices are safe enough and that attempting to further perfect them wouldn't make much if any significant difference.

Perhaps the answer is as simple as accepting that current pilots can easily work around existing system difficulties?  If the systems were somehow completely free of such annoyances, would it really matter?  Would many pilots even notice such perfection except in very unusual, perhaps artificially created training situations?

Perhaps this post is as much about philosophy as anything?

[Will]

Good questions. I think humans will probably always be needed. We're very good at adapting to new circumstances. Plus, we can generalize from past experiences, and invent new procedures if need be. That kind of stuff is very hard to program. However, we may continue to see smarter autopilots, like auto throttles that are so smart you only need two buttons: AIR and GROUND.

Here in the United States, we don't have much in the way of accidents that happen because people can't understand the airplane, but some foreign carriers are vulnerable to this kind of confusion.

I talked with a Chinese pilot the other day, a new hire for Hainan Air, who was in America for 200 hours of introductory light piston training. (He came over with no flight time whatsoever, zero hours.) He'll go back home directly into the cockpit of an A320.

What you miss with these ab initio programs is meaningful experience in stick-and-rudder airmanship. Most of these programs put people with minimal time into sophisticated airplanes with outstanding automation, and then they fly into airports that always have ILS approaches down the runway.

Contrast that with the typical American civilian pilot, who works his way up doing several thousand hours of time without automation before they ever sit in a plane with an autopilot. They know their stick-and-rudder stuff, so when the flight regime is odd or the automation gets confusing, or the ILS is broken, they are comfortable.

I believe this is what happened with the 777 at KSFO. They were cleared to fly a visual approach to a runway without ILS, and they got so wrapped up in getting the plane lined up properly that they forgot to monitor the airspeed.

Not every foreign carrier has this ab initio philosophy. But some have to, because there's no way to work yourself up through the ranks of pilot jobs in a place like Thailand or Korea, where there's no such thing as light plane commercial aviation. The American entry-level jobs of pipeline patrol, air ambulance, sightseeing tours (which is how I built my first 2000 hours), CFI work, and small scale night cargo just simply don't exist. So the pilots get hired based on exhaustive aptitude tests and then go immediately into a fully automatic airplane.

[Phil Bunch]

Will,

Thanks for your comments and insights.

IIRC, part of your airline experience involved an airliner that didn't have an autopilot - at least that way you weren't having to work around quirks in the system or second-guessing what the system would do next.

I also thought about the 777 that seems to have been hand-flown into the ground on landing as I wrote my post in this thread.  I suppose an even more automated system might have landed safely even without an ILS, but that appears to be some distance in the future since it takes so long to modify or replace such systems.  

The military seems to be headed to aircraft without pilots more rapidly than their civilian counterparts.  Their drones are flown by "pilots" who mostly have usually never flown any real aircraft, from what I've read.  The general sense is that the US Air Force has purchased its last manned fighter and bomber - it's apparently too costly and inefficient to have person(s) physically in the cockpit.  One of the many things they are working on is figuring out how to award medals for unusually effective remote piloting.  The old guard military is used to risking life and limb in order to earn medals!

I believe it will probably take a long time to persuade all the regulatory authorities to approve airliners without physically present pilots and to convince the public that it's OK to board an airliner with no in-person human pilots.  My mild discomfort at becoming a little more aware that automated flight is (inevitably) technically imperfect is much easier to deal with than trusting a fully automated, remotely-piloted airliner.  A good human pilot will work to manage a crisis if something goes wrong with the aircraft.  No pilot, no in-person human intervention.  Even though an airline accident is extraordinarily unlikely for any specific flight, especially in developed countries, I still would prefer to have an experienced, dedicated real person in the cockpit just in case something might fail.  The Hudson River ditching in New York is (IMO) an example of a situation where a human pilot was able to quickly and effectively intervene and manage a rapidly deteriorating situation.  

Or, is a remotely present human drone pilot, perhaps with no traditional pilot training, good enough for most emergencies?

Of course, some people would argue that a good set of well-programmed computers will do better than most human pilots...but how can we perfect or nearly perfect the associated hardware and software?

[Jeroen Hoppenbrouwers]

Current aviation computer doctrine is that only fully understood and fully specified cases are supported. Fuzzy problems are explicitly left out; fully deterministic everything is the Holy Grail of certification. This very principle prohibits any artificial intelligence even near a flight deck.

It is quite possible that it is exactly the drone, backed up by remote pilot, which will pave the way for AI in avionics. Not by scientific mathematical proof, but by statistics. Safety is about statistics. If the AI survives sufficient numbers of problems in flight, it can be considered safe. Probably safer than humans, statistically. We just need a million hours of drones the size of a transport aircraft, and the associated number of real problems and emergencies in flight.


Hoppie

[John Golin]

As a side note, an X37B just performed a fully automatic recently and touchdown after 675 days on orbit...

http://youtu.be/65oh_fpMwM0

A bit right of centreline tho

[Ray_CYYZ]

Can you just imagine a bunch of script kiddies getting some code  off the internet that lets them pop their PS/4 or Xbox onto the net and grab control of aircraft?

As we continue to find out on a daily basis nothing is actually secure. This blind faith in technology is pretty scary.

Even if they made it simple, would it ever be safe?

[Jeroen D]

Even if they made it simple, would it ever be safe?

Depends on your definition of safe. As Hoppie pointed out, if you go by statistics, I'm sure it will look great from a safety point of view in the very near future.

To be honest, I must admit I do feel for the statistical approach. Properly done of course.

Just as an example, different industry. Look at Formulea 1. Nearly two decades without a fatal/serious accident. Now everybody goes completey overboard and demands more safety. But this is a inherently dangerous sport.

The same of course is true for aviation. It is the safest mode of transportation. We could save more lives and serious injuries to make DIY work at home more safe than to try and improve aviation safey.

Unfortunately, some of this is very much driven by the media. You electrocute yourself at home doing some DIY work is not going to get you on the front page. Even though it happens thousands of times a year worldwide. One plane vanishes and we have the internet exploding with complot theories and people "demanding" more transparency, safety whatever.

I feel for those who lost a loved one in these crashes. At the same time, you have to ask yourself if people who are party to a (human) disaster are the most relevant to listen to on what should be done. You could argue that they are the only ones you should listen to, as they bring these accidents to a very personal level. At the same time, the statistics show a very different pattern.

Jeroen

[torrence]

As a side note, an X37B just performed a fully automatic recently and touchdown after 675 days on orbit...

http://youtu.be/65oh_fpMwM0

A bit right of centreline tho

Interestingly, the Russians demonstrated this capability back in '88 with the Buran.  Also, I think a very nice automated landing, as I remember.  The Russians never continued with this type of manned spaceflight for a number of complicated Geo-political and technical reasons, but Buran was a pretty impressive program.  I saw the craft, and its An-225 carrier at a Paris Air Show back in that era.  Actually, I remember being most impressed with the Anatov - wow!  I thought the C5A was big.

Wouldn't be surprised if they haven't got something like the X37B, or will have.  

And we're still paying them to take US and international crew to the ISS with older technology.

Cheers,
Torrence