Years ago, I flew an autoland in a Boeing 777 simulator. It was at the end of a two-plus-hour session in what was then a cutting-edge plane—it’s still pretty remarkable. In the dark, we flew toward a landing at Dulles International Airport in Washington, D.C. Well, to be specific, we watched the airplane fly itself. Now, the idea of monitoring a plane while it’s on autopilot is no big deal to pilots. We don’t even think twice about it. But keeping an eye on a plane that’s landing itself? Well, that’s a whole different story.
So Garmin’s Autoland isn’t breaking new ground in being able to land the plane, though there are very few planes that can do it. But those planes are much more limited in what they can do. The 777, for instance, needs a Cat II ILS to do it—not to mention two ATPs and two of everything else, for that matter, including autopilots and ILS receivers.
The Piper M600 SLS with Autoland can do it with no pilot at all. And it can autoland at any airport with a GPS approach with vertical and lateral nav, of which there are many thousands in the United States alone.
Still, when it comes to the science, one of the most fascinating elements of Autoland is how the system figures out where to land.
The logic is no secret. It’s looking for an airport that’s got the aforementioned type of approach, that’s long and wide enough to land the M600 on (with some fudge factor), and that’s quick to get to, and with good weather. The fact that it can make this call in three-tenths of a second, about the time it takes a world-class sprinter to react to the starter’s gun, is hard for me to fathom.
While I was in Olathe at Garmin, I met with program manager Bailey Scheel, senior software engineer Eric Tran and aviation systems team leader Ben Patel and got to talk about how they made Autoland think like pilots think. The process involved creating algorithms that weigh different factors, such as weather en route and at the airport, differently. Remember that the system integrates with Emergency Descent Mode (EDM) and ESP envelope protection before it even gets started. So there are complex decision trees built into the system before it even activates.
Once it does start doing its thing and begins the search for a diversion airport, it factors in all of the same factors human pilots do when making that same decision.
So the question becomes, how does Garmin get Autoland to perform like a pilot operating at light speed (literally) and with immediate sources of data for every question that springs to mind, like what do I do now that the one runway with a suitable approach at this nearby airport has a tailwind component? We’d fret about it, even if just for a moment. (There is, for the record, no fretting algorithm built into Autoland.) The engineers started by asking themselves how pilots do it. In the process, they came to some remarkable realizations about how aeronautical decision-making works. The results speak for themselves.
And while it wasn’t specifically relevant to my flight that day, we did discuss the lessons learned in undertaking such an ambitious project. I asked if, at some point, we might want to teach new pilots to think through such processes the same way Garmin’s software does.
Then again, we already do, in a way. It’s just that our teaching of it in most cases isn’t well structured. The process, which we call “experience,” pretty much leaves such learning up to the pilot and spreads the lessons out over the course of years or even decades. Maybe we can do better than that.
Are you an aviation enthusiast or pilot? Sign up for our newsletter, full of tips, reviews and more!