Flying an airplane

Do pilots actually fly commercial airplanes or is it automated (as I’ve heard)?
If it is automated, to what extent is the automation?

Yes pilots do fly the airplanes. Typically the takeoff up to a certain altitude is hand-flown (no autopilot) and then the autopilot is engaged. The altitude varies based on the airline’s procedures, type of aircraft, flight conditions, and pilot’s mood. On most airline flights you can expect about 80% of the flight to be flown with the autopilot engaged. At all times when the autopilot is engaged the pilots must monitor it to be assured it is flying the desired path. The pilot flying directs the autopilot to perform certain functions through several different controls, knobs, and even computers. At any time the autopilot does not perform properly it can be disengaged and in fact it has to be in that situation. During the approach and landing phase the autopilot must be disengaged on most airplanes at a certain altitude as well. The question is really best suited to each individual pilot and his/her preference to hand-fly more or less. Pilots are given latitude as to how much or little hand-flying is done on a given flight as long as the airplane is operated within certain standards. It is always a good idea to utilize the automation to assist and relieve workload. It is also important not to become totally reliant on the automation or complacent in duties while operating the flight. The level of automation can be taken up or down several levels and does not necessarily mean turning the autopilot on or off. For example, you can elect to fly without the autopilot but with the assistance of a flight director. A flight director is just what it sounds like, it prompts the pilot to manuever the airplane in order to get the desired results. Each airplane has different abilities from no autopilot at all to an airplane that can land itself (yes this happens with people on board). In any case, airline pilots are trained to proficiency without any of this automation in the event of system failures so safety is not compromised.

This is correct, but flight crews seem to vary in just how closely they watch things. I was on the flight deck of a British Airways 747 chatting with the pilot somewhere over the North Atlantic when the autopilot “decoupled” for an unknown reason. This produced a more abrupt maneuver than the flight attendants (who were serving dinner) would have wanted. None of the flight crew was paying any attention at all when this happened. (About 5 seconds later, they were paying close attention.)

And welcome to the SDMB, both of you.

Wow, two first posts in the same thread! Welcome!

dlzdude did a nice job of covering what’s involved.

Just to add a little info:

Most commercial airliners today have the capability of “autoland” which is just what it sounds like - the airplane lands itself. The only time this is required is when the weather (visibility) is really bad, or if the airplane needs to do a landing (airplanes can have currency requirements for landings). When the weather is bad this helps out, with us monitoring the airplane and looking outside. When the airplane requires an autoland for currency we’ll grumble and see how long it can be deferred: no pilot wants to let the airplane steal a landing!

We can also be dispatched with an inoperative autopilot. There are numerous restrictions associated with this (ie no LA - Sydney while hand-flying it all the way), but it does happen.

Xema any sort of autopilot disengagement at cruise would be an unexpected event. Most airplanes use different trim motors for the autopilot and manual trim. If the autopilot is trimming a lot (say, for an airspeed change) and gets disconnected it can produce an abrupt pitch moment. To prevent this we “guard” the autopilot when we disconnect it. This means that we hold the yoke to prevent any unexpected movement.

The point of all this is that the crew you were talking to WAS paying attention to what they were supposed to: they knew where they were, they were listening to the radio, they knew how much fuel they had, etc. To prevent the jolt from an unexpected autopilot disengagement you’d have to spend the entire flight with your hands on the yoke bracing for the instant it happened. If you have to do that, why even have an autopilot?

Sorry, did that sound defensive? :slight_smile:

Really, if the autopilot disconnects at cruise it will catch 999 crews out of 1,000 by surprise. Whether or not it is felt by the passengers depends on how well the autopilot had the airplane trimmed up at the time of disconnect.

Oh, and for the OP:

I hand-fly the departure as long as I can. In nice weather this means all the way up to cruise altitude. In crowded airspace or with lousy weather I hook up the autopilot earlier to relieve the workload so I can concentrate on other things.

The same goes for descent: in nice weather on my way to a not-too-busy airport (i.e. Tucson, Tampa) I’ll hand fly it from 5,000 feet on down. On a nasty day into La Guardia once again I let the airplane do the grunt work while I concentrate on other things and disconnect the AP at anywhere from 500 - 1,000 feet.

I agree that there is no practical way to monitor things so closely that this would be prevented. But in the incident I described, the change was dramatic – one moment the closest guy to any instrument or control was the pilot, about 6 ft away, facing aft and chatting casually. In just a few seconds, a lot more attention was being paid.

OK, that does sound a little less vigilant!

If you want to read about a truly scary autopilot disengagement go here.

Yikes!!!

(Hint: it’s an Aeroflot A310 crash from 1994).

At the risk of writing an essay, I’m going to offer some amplification on the good answers to the OP’s question. What’s missing is the context. And that leaves, I think, a very erroneous impression of the overall situation.

The pilot is ALWAYS flying the jet. At least I always am, as I’m sure the other pros on the board do as well. The only difference is how he/she interacts with the machine. That point is often lost in explanations to laymen and does a disservice to both pilots and piloting and to the public’s understanding.

So here goes …
At the fully un-automated extreme, the pilot has his/her hands on the yoke (steering wheel) & throttles and directly controls the pitch angle (nose up/down), bank angle, and power by hand.

The sum of those three settings drives the aircraft up/down, left/right, and fast/slow versus where the pilot wants to go. And every bump in the air knocks the aircraft slightly off the intended path and so constant attention and continuous control corrections are required.

Just like in driving, where you’re constantly adjusting the wheel to stay in your lane and adjusting the gas pedal to stay on-speed. We use tighter tolerances than you might while driving, so the workload is higher and we’re dealing with 3 dimensions, not 2. And at 200 to 600 mph, a little error grows into a big one pretty quickly. So hand-flying can absorb a significant chunk of the available processing cycles in your brain, particularly in turbulent conditions.

All ground taxiing and all takeoffs and most approaches and landings are done this way. Figure the first 3 minutes in the air and the last 10 minutes in the air of 99% of airline flights are “hand-flown” as we say.
The next level of automation is when the pilot engages an autopilot and/or autothrottle. Those computers simply maintain a fixed pitch or bank angle and a fixed power setting. The pilot then controls the aircraft by turning small knobs or thumbwheels on a control panel to adjust the desired pitch or rate of climb and desired turning or going straight and desired speed or power.

At this point the pilot is still very actively controlling the aircraft, it’s just mediated through a couple of computers to make the flight smoother and leave the pilot a little more spare mental energy for strategic thought. The computers compensate for the never-ending small bumps which constantly pull the aircraft away from the desired behavior.

The effect is similar to you controlling the speed of your car with the cruise control buttons on the steering wheel instead of the brake and gas pedals. You’re still actively in control, but the car will handle speed control on hills so you don’t have to compensate for the slope, and all speed changes are done by your thumbs, not your feet. You’re still driving the car in all respects.

This was pretty much as fancy as autopilots got into the middle 1970s. Even some older airliners flown today can do little more than this.

In this mode of automation, the pilot is actively doing all strategic thought (“where am I going?”) and most of the tactical thought (“how am I going to get there?”). The computers are taking care of the nitty-gritty details (95% RPM or 95.5% RPM? 3 degrees nose up pitch or 2.7 degrees?)

This mode of automation is generally used for most initial climbs and most arrival descents. The situation is fluid enough that the pilots want very direct control of the jet, but they don’t need the instant control that comes with hand-flying; neither do they need the burden of sweating the tiny details. If the air is smooth but the seatbelt sign is on, it’s a good bet this is the mode in use.
The next level of automated control is to attach a navigation computer to the autopilot/autothrottle. These were invented in the late 1970s and became commonplace in the late 1980s. All modern airliners are so equipped, although many carrier’s older planes still don’t have this feature.

The nav computer (generally called a “flight management system”, or “FMS”) has a keyboard and screen and permits the pilots to enter the intended route from here to there and intended altitudes along the way. This is called the “flight plan.” It is normally typed into the FMS while still parked at the gate before departure.

The FMS is connected to some system that can determine the aircraft’s position at all times, such as gyroscopic Inertial Reference Systems ("IRS"s) and/or GPS receivers. It’s also connected to the speed and altitude instruments and compasses, etc., so it knows the aircraft’s flight situation in detail.

So once airborne the FMS knows where it actually is in space and, from the flight plan, it knows where it’s supposed to be and where it’s supposed to be going.

The third level of automation is to connect the two, so now the FMS is driving the autopilot. The FMS knows how steep to climb, how fast to go, how much power to pull from the engines, where it is and when to turn left at Bakersfield to get to the next point on the route.

Most time spent in level cruise flight is spent in this mode. Likewise, the upper parts of the initial climb and descent are normally flown this way. This offloads the pilots to deal with strategic issues. And, frankly, it’s damn hard to fly precisely for hours at a time droning across the sky. The FMS’s attention never wanders, barring the occasional malfuntion.

So now the pilots control the aircraft by typing commands into the FMS. For example, imagine we’re over the Colorado Rockies going to Chicago O’Hare, and the intended route must change due to traffic or thunderstorms ahead.

The pilots consult with ATC to determine a mutually acceptable plan of action, then the pilots will enter into the FMC something like: “delete the future part of our flight plan from Denver to Hayes, Kansas, to Wichita, to Kansas City. Replace that with: from right here go to Pueblo, Colorado, to Liberal Kansas, to Kansas City, then pick up the rest of the route into Chicago as previously programmed.”

Naturally it’s all entered into the computer in mumbo-jumbo that’s quick to type and would make any DOS programmer proud, but the concept is as I’ve described.

And once the pilots type all that up and press [OK], the FMS tells the autopilot to turn right and sure enough, it goes to Pueblo and so on.

So the pilots are still actively controlling the aircraft, but two steps of indirection removed from actually touching the yoke & throttles. The pilots are still exercising strategic control but have delegated tactical control to the FMS which has delegated nitty-gritty control to the autopilot.

So who’s “flying” the plane? The way I see it, I’m flying the plane via the FMS.

And that’s as automated as it gets. On a typical 3 hour flight, the FMS plan will need to be amended about 20 times; 3 to 5 in climb, 3 to 5 in cruise, and 10+ during the arrival. For a bad weather day, or horrendous traffic, those numbers can quadruple. There’s a now-old joke in aviation that the pilot has forgotten how to hand-fly but can now type 75 words per minute. That’s entirely a comment about the workload of using an FMS to steer a plane.

There are efforts underway to connect the pilot’s FMSs with ATC’s computers so the mutually agreed plan can be datalinked up to the plane, thereby reducing the potential for confusion or typos. But even then, the negotation of what to do will be between human pilots and human ATC controllers, and both sides must accept the plan change before it gets executed by their respective computers.
Lastly, a few words about automatic landings. For the pros, you’ll see I’m going to simplify the heck out of this, but I think the gist remains both correct enough and relevant to the public’s viewpoint.

Here goes …

Most runways used by airliners are equipped with a special radio beacon called an Instrument Landing System or “ILS”. Suitable receivers on the aircraft can receive the beacon and display the path it represents to the pilots. Very crudely put, the ILS transmits a very narrow funnel which ends on the runway where you’re supposed to touch down. From there the funnel extends upwards and outwards back along the approach route for about 20 miles.

When hand-flying, the pilots will manuever to cross into the ILS funnel 5 to 15 miles away from the airport and then turn towards the airport. Pointers of some sort on the intruments will then show the aircraft’s position up/down and left/right compared to the centerline of the funnel. The pilots steer the jet to get centered in the funnel and remain centered as they slide down the ever narrowing funnel towards the runway.

Eventually they get close to the ground and (normally) come out the bottom of the clouds, see the runway and land visually.

About 95% of all airliner approaches / landings in good weather down to poor-but-not-awful weather are flown this way; entirely by hand. Even on bright sunny days we’ll use the ILS to “calibrate our eyeballs” and ensure we’re exactly aligned vertically and horizontally as we descend towards the runway.

For each runway there’s a specific minimum altitude to follow the funnel down to. If you come out of the clouds first, you see the runway and land. If you get to the minimum altitude and still don’t see the runway, you immediately “go around”, i.e. climb back up away from the ground and then try plan B, whatever that may be. (And rest assured we have backup plans running out to Z and beyond.)

That minimum altitude varies between 500 and 200 feet above the ground depending on the surroundiing terrain, quailty of the runway, equipment on the aircraft, etc.

Meanwhile, while following the funnel those last few miles to the airport, the pilots are slowing down from manuevering speed of, say, 250 knots to landing speed of, say 130 knots (itself a non-trivial process), deploying landing flaps, landing gear, talking with various ATC people, tracking where they are versus various distance/altitude checkpoints along the route, and a hundred other things. It’s a LOT busier than changing a few lanes to get off the freeway in moderate traffic. And none of those activities is automated or automatable at the current state of the art.

That’s the normal, or hand-flown approach.
Now, for an “automated” approach, we connect the ILS up to the autopilot. After we manuever into the funnel it will now steer down the center of the funnel. That’s it.

The pilots still slow the aircraft and do the other hundred things. All the computer is doing is following the funnel. That’s an important task, and having the computer do it is a great workload reducer, but the airplane is hardly “flying itself.”

After all the ancillary details are done and we’re getting within a minute or so of the ground, the pilots will be watching everything closely and when either the ground hoves into view or the minimum altitude is reached, they will turn off the autopilot, take over manually, and either land or go around as approriate.

Because the autopilot can, statistically, fly the funnel more tightly than human pilots, the minimum altitude for seeing the runway on an automated approach can be reduced to as little as 100 feet above the ground, or just 7 seconds before impact.

I say “statistically” because on a good day a good human pilot can outperform the older rattier autopilots and equal the latest geewhiz ones. But the computers seldom have a bad day; they either work, or not. The standard deviation of human performance is much larger.

So that’s the “automated” approach. Lots and lots of human work, some human tension, and a smidgen of machine assistance. And about 1% of all approaches are done this way, usually when it’s very foggy or snowing heavily. We also do them periodically on nice days to practice the procedures and test the equipment.
Finally, the “automated” landing. We connect one more gizmo into the autopilot, the “radar altimiter”, or “RA.” That device measures the distance from the wheels to the ground directly below to an accuracy of a couple of feet.

So now the autopilot follows the funnel and when the RA says we’re about 30 feet from touchdown it tells the autopilot to retard the power and raise the nose a smidgen, then wait until the wheels touch the ground. That’s it.

The pilots still slow the aircraft and do the other hundred things. And watch this process very closely. For these automated landings there’s no requirement to see the ground at all before touchdown. We’ll generally see a runway light or two peeking through the fog just at touchdown or maybe an instant before. If the airplane’s not in the right place, there’s no time to do anything about it.

The pilots then manually slow the airpane down to taxi speed while the autopilot keeps it rolling down the middle of the runway. Then we turn off the autopilot and try to feel our way off the runway and to the terminal with visibilities as low as a 1/10th of a mile, i.e. only a couple of airplane-lengths.

In places like Europe prone to long periods of dense fog, we can often land in conditions were we can’t see well enough to taxi safely. Airplanes are often towed from the runway to the gate by local ground workers who know the airport perfectly and are sitting only a couple feet above the cement, rather than 2 to 3 stories above the ground as we are.

So the “automated landing” is more automated, but it still requires a goodly amount of human effort and training and talent (and sweat) to get in position, get it set up, and monitor its every move down to the climax at touchdown while still manually doing the other hundred things I mentioned. And then instantly at touchdown, we spring into action to slow the aircarft down and begin manually driving to the terminal.
Bottom line for this quasi-rant:

Anyone who thinks pilots are snoozing while some computer on the ground or in the air is doing all the work is sadly misinformed. Computers are handy helpers, but our role today is that of very active management of the aircraft and its computers.

And the idea a flight could be 100% pre-programmed and then played out like a tape recording is nuts. No flight has every operated 100% as per the original plan; something always changes and something always comes up which needs a real-time reaction from the crew.

There will be more automation in the future, but it’ll continue to nibble at the edges.

Reminds me of something I heard on the tube awhile back;
“The technology’s ready, the pilots’ are not.”
They were talking about UPS totally automated cargo flight studies, IIRC.
Peace,
mangeorge

Well written LSLGuy, I’ve often wondered what’s going on up there in the cockpit and you’ve done a wonderful job explaining it to the vast majority, the passengers! Thanks :slight_smile:

Actually, you may have educated one or two of us “amateur” pilots who haven’t had much experience with autopilots.

(My last encounter with an autopilot involved an autopilot that was supposed to be disconnected for previous misbehavior. It was… educational. But off topic to flying commercial airliners)

Good explanation LSLGuy and welcome to the boards Lostris269.

:smiley:

Wow, these explainations make me happy.
I generally hate flying on planes, but I think I just need to be on a plane with a SD pilot and I’d be a happy camper! :slight_smile: