October 2006 Volume 49 / Number 10
Bruce Landsberg has been executive director of the AOPA Air Safety Foundation for the past 14 years.
As we move into increasingly sophisticated aircraft with more data, more glass, and more speed, what role should an autopilot (AP) play? You wouldn't think that a great laborsaving device like the autopilot could create such a diversity of opinions.
Autopilots have been around for nearly the entire history of powered flight, starting with a primitive model invented by Elmer Sperry just before World War I. Today, autopilots are essential on turbine aircraft, commonplace on GA cross-country aircraft, and even showing up on trainers. How should pilots use them and how much should we depend upon them?
There are those who live by AP: wheels up, AP on; wheels down, AP off, or, if the ceiling is low, the pilot-engagement point becomes "runway in sight." If the gear is fixed, 500 feet above ground substitutes as the on-off point. Those who routinely fly this way are abdicating their basic flying skills to the hardware.
The other extreme is the pilot who hand flies a four-hour trip even though automation is available. He or she may be highly proficient, but the wear and tear on the pilot is likely to be significant, depending on the physical condition, age, and experience of the marathoner. Such a pilot's flight path may be somewhat erratic, especially if the avionics require much attention, and flight management may occasionally become secondary to aircraft control.
When I started flying, back in the Pleistocene era, auto-pilots were a rarity in light aircraft. New instrument pilots had to keep all the balls in the air manually, and while it may not have always been graceful, it worked — mostly. We weren't trained in the use of automation and even when it was available, many of us considered it a crutch.
After checking out in bigger, faster aircraft with more sophisticated avionics, I learned that flight management was more than just a term; it was a useful concept for negotiating high-density airspace and dealing with inevitable changes, allowing the aircraft to perform the basic flight tasks, with supervision. The autopilot became an essential crewmember in the single-pilot cockpit.
As a new IFR pilot, I checked out in a Piper Arrow. The avionics packages in those days were mostly basic and standardized, and the instructor provided a cursory briefing on the autopilot. It was a single-axis device that held heading or track, but did not have an altitude-hold function.
On one of my first Arrow trips in the clouds, I was hand flying, minding my own business, and marveling at my ability to survive. New York Center decided that things were going entirely too smoothly, and gave me the obligatory reroute.
I copied the clearance, which included intersection names known only to FAA planners, while attempting to hold straight and level, retune the VOR, reset the OBS, and fumble with the transponder. On most tests, getting four out of five elements isn't bad and 80 percent is considered a passing grade most everywhere. This overlooks my F grade for altitude control. At some point during the breakdance the controller politely reminded me to check the altimeter setting, probably knowing full well what was going on. The autopilot, even without altitude hold, would have been a fabulous asset if I had used it, and the whole deal would have been accomplished far more elegantly.
We are required on the instrument practical test to handle the aircraft, reroute, reprogram, and stay within checkride tolerances. In the real world, while dealing with turbulence, fatigue, passenger distractions, and myriad other items, the reality is that those tolerances are sometimes stretched into "pink-slip" territory. It shouldn't happen, but then IFR life isn't exactly as portrayed in training. Ask me sometime why that's nearly impossible to do effectively. Also understand that I'm not advocating loosening the standards.
When Cessna introduced the Citation line of jets, the autopilot was integral to the FAA's single-pilot approval. Pilots were taught that the autopilot flew the aircraft and it was to be used in all normal circumstances. If the autopilot broke before takeoff, the flight was canceled and if it failed en route that was an abnormal procedure. The pilot was expected to be able to handle the aircraft, but it was appropriate and expected to ask for ATC assistance, if needed, and divert to the nearest suitable airport. The extra time and mental processing power was used to manage the avionics and stay at least 10 miles ahead of the aircraft. Required autopilot use was a major attitude shift for the light-aircraft crowd, most of who used autopilots periodically and didn't really trust them. With the arrival of very light jets, you can be certain that the autopilot will be a required piece of equipment and that pilots will be expected to use it religiously.
Let me clarify something for the "what-if-it-fails" crowd. In searching the AOPA Air Safety Foundation's accident database back to 1983, we were unable to find a single accident where NTSB considered autopilot failure as the probable cause. That doesn't mean that it can't happen or that you should give up practicing hand-flying skills. But a shift in attitude is appropriate as we transition to complex avionics packages that deliver a far better flight-path product and situational awareness, but need much more programming and demand more attention than the old ones.
Today's autopilots are much more reliable than the humans programming them. I concede that a few units built on Friday afternoons before a holiday may be less trustworthy than a politician at a PAC reception. But as a group, humans are far more likely than our electromechanical helpers to deviate from a heading, miss an altitude, blow through a final approach course, or wobble down the localizer.
As the equipment has changed, so too should the testing and training for light aircraft, mirroring the single-pilot jet that our cockpits are now emulating. How about treating autopilot failure in actual instrument conditions in the same way as an engine malfunction on a multi-engine aircraft? When an engine acts up or fails, the mode of operation changes. We set priorities very carefully, advise ATC that we have a problem that may develop into an emergency, or declare the emergency outright. A diversion should be made to the nearest suitable airport, not necessarily the nearest airport, which may have a complex and difficult approach. Vectors-to-final is a smart way to handle this.
I have played and inflicted "what-if" games on students, including scenarios when a controller's radar is down, or a position 300 miles from nowhere, and a bogeyman that jumps out if you don't keep it all going perfectly. The traditional approach to training for an autopilot failure is often to continue the trip as if nothing had happened. Just suck it up, son, and, by the way, I've got a reroute for you with a hold and a course reversal on a back-course approach with a dogleg at the final approach fix. It's a par four, and mind the sand trap. The proper answer in the real world is, "Unable — we've got an equipment failure and I'll need vectors to the ILS at downtown municipal." If you're sinking, ask ATC for the localizer frequency, inbound course, and altitudes in sequence so the workload remains manageable.
If you can reasonably handle a bit more, that's good, but in the real world, the idea is to manage risk and work load. In the minds of some instructors, pilots should just meekly accept all the stuff that is shoveled into the cockpit by the CFI or ATC, instead of acting as pilot in command. Here's what's bad about this 400-pound bench-press approach to training: In an actual situation, pilots tend to react as they have been trained. If the autopilot dies, they may revert to the Superman mode they learned in training, even though most of us aren't Superman or Wonder Woman six months after training. In really nasty weather after a series of long workdays and perhaps not flying quite as much as we would have liked, it's not going to be the same as the environment created by the tough-as-nails coach omnipotently sitting in the right seat. To continue the sports metaphor, this isn't the time for the Hail-Mary pass — just get first downs until you get to the runway. If you can only bench-press 100 pounds, train to where you might get to 120.
So how do you balance being a good enough hands-on pilot in command with intelligent use of an autopilot? I like to hand fly departures until about 5,000 feet, let the machinery do the mindless en route part, and, on at least every other trip, hand fly the approach. This keeps me conditioned, but I also go for weightlifting sessions and coaching every six months.
Some flight schools are now buying full-glass-cockpit aircraft without autopilots to save money and perhaps to train pseudo supermen and women. It's a false economy and premise. Either go with the full package and learn how to use automation intelligently or stick to steam gauges and basic avionics. Let's train for the real world!