Following is a transcript of the video.
Narrator: Autopilot isn’t as “auto” as you might think. There’s no robot that sits in the pilot seat and mashes buttons while the real pilot takes a nap. It’s just a flight-control system that allows a pilot to fly an airplane without continuous hands-on control.
Basically, it lets a pilot fly from New York to Los Angeles without white-knuckling the controls for six straight hours. But how does it actually work? Kind of like a polar bear. A polar bear’s core temperature sits at about 98.6 degrees Fahrenheit. It is so well insulated against the frigid Arctic cold that it often overheats. When that happens, its body reacts by releasing excess heat through its hairless parts, like its nose, ears, and feet. The polar bear’s body temperature returns to a comfortable 98.6, and it’s free to hunt seals another day. That cycle is called a negative feedback loop, and it’s the same way an autopilot functions.
A negative feedback loop is a self-regulating system that reacts to feedback in a way that maintains equilibrium. Generally, it uses a sensor to receive some sort of data or input, and the system uses that data to keep functioning in a preset way.
For the polar bear, that preset is body temperature. For an airplane, it’s lateral and vertical movement. A modern automatic flight-control system is made of three main parts: a flight-monitoring computer, several high-speed processors, and a series of sensors placed on different parts of the plane. The sensors collect data from the entire plane and send them to the processors, which in turn tell the computer what’s what.
AFCSs come in three different levels of complexity. There are single-, two-, and three-axis autopilots, based on the number of parts they control. Single-axis controls the ailerons, which are these guys. They make the plane do this. Single-axis autopilot is also called the “wing leveler” because it controls the roll of the plane and keeps the wings perpendicular to the ground. Two-axis handles everything the single-axis does, along with the elevators, located here. They move the plane like this. And three-axis juggles those two plus the rudder. That one there is in charge of this movement. Then the computer tells the servomechanism units what to do. Servos are the little instruments that actually move the parts. All of these pieces come together to make sure your plane stays in the air, where it belongs. But they don’t just work on their own.
The success of the autopilot depends on the knowledge of the actual human pilot.
Greg Zahornacky: Autopilots are dumb and dutiful, meaning this: that if you program them incorrectly, they will kill you.
Narrator: Dumb and dutiful are the “two Ds of automation,” according to Earl Wiener, a former US Air Force pilot and an aviation scholar. He once described autopilot as, “Dumb in the sense that it will readily accept illogical input; dutiful in the sense that the computer will attempt to fly whatever is put in.” It’s crucial, and I cannot emphasize this enough, that you know how to fly a plane before you use an autopilot. Step one is inputting a flight plan. And step one is also where things could start going wrong.
To get from New York to LA, a pilot needs a route. That route translates to a flight plan, and that flight plan gets punched into the computer and logged into the database. If the pilot doesn’t know what the heck they’re doing, then they can end up programming the autopilot to fly the plane upside down or to spell out “I’m a Bad Pilot” in the sky. If they correctly navigate step one, step two is simply turning on the autopilot. The system executes the flight plan and takes over from there.
Zahornacky: That will stay operational until such time as they tell it or turn it off. But it is capable of flying the aircraft essentially from takeoff all the way to touchdown and including touchdown.
Narrator: But you can’t just tap it and nap it. It’s the ABCs of autopilots: Always be checking. Because autopilots can and do fail. Sometimes it’s user error when entering the flight plan. Sometimes it’s a sensor or servo malfunction. Either way, this is when it becomes very important that an inflatable toy isn’t flying the plane.
– Why is it doing that?!
Zahornacky: If it’s not doing what I expect it to do, I’m gonna disengage the autopilot. I’m gonna go back to hand-flying the aircraft and say, OK, this is what I want you to do. I’m gonna rebuild it again.
Narrator: The good news is autopilot will never take over a plane, à la HAL. Worst case, the pilot turns it off and on again or pulls the circuit breaker if that doesn’t work and reprograms it to behave itself. Worst-worst case, the pilot just has to fly the plane themselves.
Zahornacky: So, I am a very large proponent of hand-flying that airplane to keep your skills high because, you know what, you’ve gotta go through a check ride at least once a year.
Narrator: A check ride is a practical test regulated by the Federal Aviation Administration that US pilots must pass to get their licenses. And most airlines require yearly check rides to make sure their pilots can actually fly.
Zahornacky: ‘Cause if it’s on autopilot all the time, how can you keep your skills sharp?
Narrator: There’s a reason we still have pilots flying planes and haven’t handed the yoke over to robots. As advanced as the technology is, an autopilot is not auto enough to think for itself, which means it’s not smart enough to fly a plane by itself, and that’s another thing autopilots have in common with polar bears.
EDITOR’S NOTE: This video was originally published in October 2019.