Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
AI Transportation

Would You Trust a Self-Landing Plane? (aopa.org) 94

"Lots of hay has been made over self-driving, self-parking, and summon features of new cars," writes Shotgun (Slashdot reader #30,919). "Now Garmin has a system that will land your plane for you."

He shares this new report from the Aircraft Owners and Pilots Association: During this first-ever autoland for me, nothing felt terribly out of the ordinary until the final 100 feet or so when it finally sunk in that in a few seconds we would be hitting the pavement and no one had their hands on the yoke.

But just as briefed, the airplane "decrabbed" from the 10-knot left crosswind and soon plunked us down just left of center on Runway 18 at New Century AirCenter in Olathe, Kansas, and then quickly tracked us back to the centerline. A few seconds later we rolled to a stop... Such was my introduction to Garmin's new autoland system, a first in general aviation...

While the notion of an autoland system seems futuristic, Garmin has been quietly working on it since 2011 and conducted its first autoland in a Columbia piston-powered airplane in 2014 and first briefed the FAA on it in 2015. Since then it has conducted some 800 autolands... Someday, with more experience and the FAA more comfortable with the notion, perhaps Autoland will be available on any given flight where the pilot doesn't feel up to the landing. Meanwhile, it's a great safety feature. "This will save lives," said Ron Gunnarson, Piper's vice president of sales, marketing, and customer support. Once certified, the system will be standard on the 2020 M600 SLS models... Cirrus plans to offer Autoland as a standard feature on the 2020 models.

The article describes it as "a digital parachute" for the pilot and passengers in emergency situations. And the original submission argues that it brings to general aviation a safety feature that the big airliners have already had for years.
This discussion has been archived. No new comments can be posted.

Would You Trust a Self-Landing Plane?

Comments Filter:
  • Huh? (Score:5, Informative)

    by Joce640k ( 829181 ) on Sunday November 03, 2019 @12:38PM (#59375408) Homepage

    Passenger aircraft have been auto-landing for decades now. If you've flown more than three of four times in your life you've probably done an autolanding.

    Passenger aircraft can actually lose their certification of airworthiness if they don't do a minimum number of autolandings every month.

    • ... it brings to general aviation a safety feature that the big airliners have already had for years

      Oh, wait. Read to the end of the summary before posting.

      (but any decent article editor would have put that info in the first line)

      • (but any decent article editor would have put that info in the first line)

        Having an editor whose job is to review a story before posting? What a quaint concept.

      • But it answers the headline question of whether we would trust it -- millions do, every day.

        • Reminds me of the old aviation problem report joke.
          Problem: auto land rough.
          Solution: aircraft not fitted with auto land. :)

      • Actually, every plane (and balloon, glider, zeppelin, etc) ever built is self-landing. You just can't self-land them more than the one time.
      • Garmin device is probably doing it by GPS Lat, Lon and altitude. I do have several Garmin devices mostly for hiking and bicycle riding, and I know all of those devices suck when it comes to height above sea level. Height could be off by 50 feet or more.

        Commercial aircraft do it by staring down a fixed radar beam attached at the back end of the landing strip. Stay in the beam and your wheels will eventually contact the center of the runway at a given rate of decent. The only way to get it wrong is for the
    • Re:Huh? (Score:4, Interesting)

      by Known Nutter ( 988758 ) on Sunday November 03, 2019 @01:23PM (#59375556)

      If you've flown more than three of four times in your life you've probably done an autolanding.

      This article disagrees with you, stating autoland accounts for around 1% of commercial flights:

      https://www.flightdeckfriend.c... [flightdeckfriend.com]

      This guy says he's used it 10 or 20 times in 30 years flying the A320:

      https://www.quora.com/How-ofte... [quora.com]

      • This guy says he's used it 10 or 20 times in 30 years flying the A320:

        Yet this guy says he loses his pilot's license if he doesn't do at least one every 90 days:

        https://www.flightsim.com/vbfs... [flightsim.com]

        (second message)

        • On the page you linked, one person said it's company policy to do them a few times per year and another person said it's a "requirement" for their crew to do them every 90 days. I don't see any mention of losing their license.

          I do see mention of being allowed to land in zero visibility. If you're into fly to Alaska and land in zero viz, you need to practice autoland. It doesn't say whether that's a company policy or an FAA rule.

          • The pilots are reluctant to use the Autoland feature because they do not want to train their replacements.

            • by caseih ( 160668 )

              Actually, no it's more that the funnest parts of flying are take offs and landings, so pilots like to do them manually. The in between bit is boring and all flown on autopilot as a matter of course. It's a challenge and an accomplishment to grease the landings, and pilots relish that sort of thing.

              • by dgatwood ( 11270 )

                Actually, no it's more that the funnest parts of flying are take offs and landings, so pilots like to do them manually.

                And also because autoland tends to result in harder landings, which passengers don't particularly like (unless this has improved in newer versions), and because if pilots didn't do manual landings regularly, they couldn't be trusted to land the plane whenever autoland won't work (e.g. when there's too strong a crosswind). And so on.

              • by tlhIngan ( 30335 )

                Actually, no it's more that the funnest parts of flying are take offs and landings, so pilots like to do them manually. The in between bit is boring and all flown on autopilot as a matter of course. It's a challenge and an accomplishment to grease the landings, and pilots relish that sort of thing.

                Exactly. There also aren't that many aircraft that actually have true autoland - you need a Category IIIA certified ILS (autoland) with dual autopilots, dual nav receivers, and still monitor it in because if there

            • That's funny.

              I'm fairly certain they use simple PID controllers, though, not neural networks. PID controllers are very predictable, while neural networks are anything but. There isn't a lot of training on a PID controller, just three numbers to set correctly.

              • It is roughly PID controllers, however much complexity is added to deal with negative gain situations. For instance, you generally go up if you increase the noise pitch. At a certain point the wings stall, and the linear assumptions required by PID control no longer apply.

                The result are algorithms tuned to handle special cases. Originally, the adaptive control people wanted to use the technology to fly planes. I don't think anyone ever let a fully adaptive algorithm loose on a plane with no constraints or

                • Thanks for that.

                  Interesting fact - most pilots will tell you that throttle controls altitude, elevator controls airspeed. Of course they are linked, they should be coordinated to some extent, but that's the basic rule.

                • It is roughly PID controllers, however much complexity is added to deal with negative gain situations. For instance, you generally go up if you increase the noise pitch. At a certain point the wings stall, and the linear assumptions required by PID control no longer apply.

                  The result are algorithms tuned to handle special cases. Originally, the adaptive control people wanted to use the technology to fly planes. I don't think anyone ever let a fully adaptive algorithm loose on a plane with no constraints or cross-checking. Given the application, the engineers wanted confidence in the program.

                  This is from controls classes. Many professors talked about adaptive control on planes. No one cited an implementation. Someone with current expertise may know more.

                  NASA used Kalman filtering for some guidance work. That is adaptive signal processing, without the control element.

                  Any software that actually controls the airplane is tested to DO-178 level a, basically it must be proven that all variables in every branch are independently tested against requirements and some additional FAA certification requirements, including robustness testing. There can still be problems if the requirements are written incorrectly, but ultimately this software is probably the most completely tested software in the world. Boeing's MCAS was not classified as flight critical software controlling the

                • It is roughly PID controllers, however much complexity is added to deal with negative gain situations. For instance, you generally go up if you increase the noise pitch. At a certain point the wings stall, and the linear assumptions required by PID control no longer apply.

                  The result are algorithms tuned to handle special cases. Originally, the adaptive control people wanted to use the technology to fly planes. I don't think anyone ever let a fully adaptive algorithm loose on a plane with no constraints or cross-checking. Given the application, the engineers wanted confidence in the program.

                  This is from controls classes. Many professors talked about adaptive control on planes. No one cited an implementation. Someone with current expertise may know more.

                  NASA used Kalman filtering for some guidance work. That is adaptive signal processing, without the control element.

                  Oh, also I've used Kalman filtering in aviation software work. In the software I was working on, it was used to filter noisy accelerometers to report the aircraft's attitude and position to the pilot. Not actually controlling the aircraft.

      • by Jeremi ( 14640 )

        From the article:

        "The system is not like the autoland systems on airliners and a few models of business jets in that those systems require a ground-based signal--a Category III ILS--specially trained and certified crews, and dual or triple redundant autopilot systems, among other equipment. Such precise ILSs are available at only a small number of runways."

    • by Dunbal ( 464142 ) *

      If you've flown more than three of four times in your life you've probably done an autolanding.

      But if there's a real pilot in the cockpit he can be there if/when the unexpected happens. I'd rather have a human up there with me who has skin in the game, who also will lose his life, than some drone operator in some air conditioned room somewhere - or better yet, some software outsourcedand to and coded by the minimum bidder... I know for sure a human pilot will do everything he possibly can to avoid a crash. Except the very rare occasion when you get a suicidal lunatic like the doomed Germanwings fligh

      • I'd rather have a human up there with me who has skin in the game, who also will lose his life, than some drone operator in some air conditioned room somewhere

        The risk of losing your life does not increase performance, it increases stress and lowers performance.

        I'd much rather have the person with a lower stress level piloting in an emergency.

        • by Dunbal ( 464142 ) *
          If you are a professional and confident in your skills you do not perform worse under stress. That's what the training is for.
          • If you are a professional and confident in your skills you do not perform worse under stress. That's what the training is for.

            Liar liar, pants on fire!

            There are people in the world who study this shit. For you, that is a total shocker! LOL Because if you knew that such things existed as basic knowledge, you wouldn't say something so stupid.

            If people have enough confidence, they'll claim to be perfect! Duh. Training, however, does not make humans perfect. It only makes them arrogant assholes who still perform worse under stress.

    • Auto-land != ILS. Watch the video - it does way more.
    • be my guest, but the pilot is the backup. I for one would be more than willing to walk away from any drone passenger liner; and demand my money back.
    • by gtvr ( 1702650 )
      If the auto-landing mechanism fails, next thing you know the flight attendants are walking up and down the aisle, asking if anyone knows how to fly. There's an auto-pilot, but if he gets deflated and the manual inflation tube is broken, you end up with Ted Striker flying the plane and Leslie Nielsen cracking jokes.
      • Randy:
        "Excuse me, sir, there's been a little problem in the cockpit."
        Ted:
        "The cockpit? What is it?"
        Randy:
        "It's the little room at the front of the plane where the pilots sit.
        But that's not important right now."
    • "Passenger aircraft can actually lose their certification of airworthiness if they don't do a minimum number of autolandings every month."

      Can you elaborate on this? Why do they have that requirement?

  • by dschnur ( 61074 ) on Sunday November 03, 2019 @12:39PM (#59375410)
    That system is designed to land an airplane only if the pilot is incapacitated. It's not designed for regular use. As for using it during an event it's designed for, I'd be more comfortable than lawn-darting the plane by myself.
    • by Calydor ( 739835 )

      It also seems to me that auto-landing an airplane should be an easier to solve task than driving a car. You have the gyroscope and make a (radar? laser?) sensor in a known angle to determine distance to the runway. Now you're certain of the height of the plane. Airplane runways have high visibility on their lines, they have lights on the sides, and most importantly: Unless something has gone HORRIBLY wrong there is no other traffic.

  • by fustakrakich ( 1673220 ) on Sunday November 03, 2019 @12:44PM (#59375438) Journal

    Be careful that they don't confuse "autoland" with their cruise missile programming.

    The fact is that automation has made commercial aviation incredibly safe, in spite of Boeing's efforts

  • I'd trust it far more than anything from Boeing!
  • by SuperKendall ( 25149 ) on Sunday November 03, 2019 @12:47PM (#59375446)

    I feel totally comfortable with all aspects of a plane being autonomous...

    There are a fixed number of airports, that you can have highly detailed information about.

    They are supplemented heavily by beacons and things for large aircraft instrument landings, no reason smaller planes couldn't make use of that data also.

    While in there air there are not really that many obstacles to hit, and with a good enough sensor array (visual/radar) you should be able to make out anything coming and avoid it - even probably birds.

    The main thing I think an automated system would be great at is reaction time to immediate dangers - like birds that were not detected, or sudden gusts when landing. There's no reason automated control of airplanes cannot quickly be as good as the best pilots, or better in some cases since they combine so much more data more immediately.

    • by Tokolosh ( 1256448 ) on Sunday November 03, 2019 @02:05PM (#59375722)

      Future airliners will have a captain and a dog in the cockpit. The captain is there to feed the dog. The dog is there to bite the captain if he tries to touch the controls.

    • > While in there air there are not really that many obstacles to hit

      Yeah it didn't take me long at all to learn how to fly around once up in the air. I accidentally went a few hundred feet too high once and you know what happened? Nothing. Lots of empty sky.

      On the other hand, if you're a hundred feet too low during a landing approach THAT IS a problem. Landing is an order of magnitude harder than flying at cruise altitude. You have to be in just the right spot in each of three dimensions plus speed,

      • Landing is an order of magnitude harder than flying at cruise altitude. You have to be in just the right spot in each of three dimensions plus speed, plus attitude in 3 axis. So really seven dimensions.

        Exactly the sort of things that computers do better than humans.

    • by AmiMoJo ( 196126 )

      These days many accidents are due to faulty sensors. Flight crews can sometimes recognize the bad sensors and use some other source of information to fly the aircraft.

      For some reason it seems that autopilot systems often fail when sensors go bad.

  • This would have been a perfectly fine submission if not for the clickbait uninformed title.
    Autoland is a common feature on commercial airliners for well over 30 years, and we already have to trust them (from time to time).

    Now, the feature that would be new and unusual is an auto-takeoff. And that's something the question could be reasonably posed wrt., i.e. "Would you trust an auto-take off plane".

    • by ceoyoyo ( 59147 )

      The real reason auto-takeoff isn't implemented on airplanes is that takeoffs are simple enough for the pilots to do themselves in any reasonable scenario. It's not worth certifying a system to do it. Even so, in an airliner you usually turn on the autopilot pretty soon after takeoff.

      Autoland is used a lot for convenience (not "from time to time" but routinely). It's also an important safety feature, as it can land the plane in instrument only conditions better than the pilots can.

      • The real reason auto-takeoff isn't implemented on airplanes is that takeoffs are simple enough for the pilots to do themselves in any reasonable scenario.

        True story: The first time I ever flew, I did the takeoff.

        (in a Cessna...)

        And... the second time I flew we had a crash landing. No, I wasn't at the controls.

  • A car autopilot has to deal with much more in the way of random/unexpected objects near it than a plane has. Airport runways tend to be clear of pedestrians, cows, debris, ... all of which will occasionally appear at the side of a road. Yes: they have to deal with unexpected gusts of wind, rain, etc, but these are expected hazards and the autopilots can be programmed/trained to deal with them - with reaction speeds better than a typical pilot.

    Really unexpected air events are much rarer than on a road.

    • Re: (Score:2, Funny)

      by mark-t ( 151149 )

      Yes: they have to deal with unexpected gusts of wind, rain, etc, but these are expected hazards...

      One is compelled to ask how you can expect something that you had already qualified as being unexpected in the first place?

      • One is compelled to ask how you can expect something that you had already qualified as being unexpected in the first place?

        His statement is correct. The category of hazard is expected. The incidence of hazards cannot be perfectly predicted, hence "unexpected."

        • His statement is correct. The category of hazard is expected. The incidence of hazards cannot be perfectly predicted, hence "unexpected."

          ..... aaaand a demonstration. I knew all about how to handle the quote tags. An expected hazard. In this case I unexpectedly missed one and hit "submit" too quickly.

          • by mark-t ( 151149 )

            If the hazard is expected, then by definition it is *not* unexpected. If you cannot predict it, then it is not unexpected, it is unpredictable, which is not the same thing.

            For example, while I do try to predict any possible ways that code I write might be made to break while before I write it, I still *expect* to find bugs in the first draft code, regardless of what I might have otherwise anticipated.

      • by Ambvai ( 1106941 )

        The same way that I fully expect to die-- but I'll leave the predictions of the precise moment and method up to the actuaries, preferring a somewhat unexpected death. (Or maybe I'll leave it to the executioner, if it comes to that.)

      • Well, it goes with the known unknowns, as opposed to the unknown knowns and the unknown unknowns...
  • Not if Boeing made it. Really, self driving cars are a fallacy that are years away from being viable in all circumstances. Most of the time, they can slow down or stop in an emergency and won't fall out of the sky. Planes, not so lucky. I have done hardware for autonomous vehicles and robots for years, with companies that have valuations in the billions. It's a lot of bs and wishful thinking when it comes to some pretty standard "edge cases"
  • If I were purchasing/outfitting a plane these days and had the budget for one of these I would definitely go for it. And use it under most normal conditions.

    But if you are running a short final on a small runway with a 30-degree crosswind 25 gusting to 35 I am not so sure I would "trust" it. Landing a light plane in that scene is no small skill that is earned with no small amount of effort and emotionally it is hard to see how an autopilot would be able to cope. Maybe that feeling would ease over time

    • Short runways are unlikely to have a CAT-III ILS installed, let alone CAT-IIIc.

    • Good for when the pilot wants a few cocktails during the flight though...

    • and emotionally it is hard to see how an autopilot would be able to cope.

      "Emotionally??"

      • by pjt33 ( 739471 )

        I interpreted that to mean that instinctive distrust is stronger than reasoned trust. It's like volunteering to go in the wheelbarrow on the tightrope across the Niagara falls: you might believe that Blondin can take "someone" across safely, but be terrified to be that someone.

      • by hawk ( 1151 )

        It's broken hearted.

        The flight computer left it for the ailerons . . .

        hawk

    • If "autopilot" can keep a inherently unstable flying wing design such as B2, airborne. then it can definitely handle whatever landing you can throw at it. In-fact "stability assist" is always active on a B2 as a human could never hope to keep it stable with manual input
      • If a system containing both A and B can achieve C, then any system containing A and B must be able to achieve D. Because it sounds easier than C.

        False.

    • emotionally it is hard to see

      Your eyes can deceive you, don't trust them. Stretch out with your feelings.

    • I have landed many times in a C152, and I understand your comment, but what is it exactly that you thing you could do better than the autolanding system, or is it just a vague feeling that you could do it better? I'm just asking out of interest, because I think that in general humans make terrible drivers and terrible pilots. I believe there would be fewer crashes in both cars and planes if we automated everything. Just my opinion as a programmer and a driver/pilot.
  • The obstacles are few and far between. The landing strips are already assisted by a several bits of simple yet powerful assisting tech that reliably worked for decades.

    Why would this be considered more futuristic than a self-driving car?

  • JFK tower to BA aircraft, just landed:
    "Speedbird 10, you landed left of centerline there."

    BA captain:
    "Correct, my good man, and my co-pilot landed to the right."

  • This really isn't a new technology. Lockheed was doing it with their L-1011 TriStar thta was widely used by TWA. It didn't contact the control tower, but it would autoland the plan.

    https://en.wikipedia.org/wiki/... [wikipedia.org]

  • Autopilot worked fine for deorbiting and gliding a Space Shuttle to a landing on Earth, assuming no damage to the orbiter itself.
  • by djbckr ( 673156 )
    Yes, Yes I would. I am a licensed pilot.
  • Computers are much better than people at solving differential equations. In fact, many types of aircraft are kept in the air by automatic control systems that a human couldn't replace.
  • An Arduino can land a scale airplane. Why couldn't a professional computer land a real one? Shit, an Arduino could probably do that, too, though it wouldn't be my first choice.

    • The Arduino would not pass the hardware and software requirements. Besides you cant compare a model airplane with a big airplane. They work and behave differently. Could an Arduino do it in theory, yes, definitely.

  • Would I trust it? Not at all.
    But as a safety backup, as long as it's not fighting a living pilot (cf the latest Boeing issue) it sounds invaluable.

  • Your pilot should be there for perverse instantiation situations like the Boing 767 Max accidents but should have full override capabilities. Most planes have fly-by-wire anyway so the pilot's actions can be analyzed for machine-learning. The one thing we have to be careful of is that the machine-learning is also used to train new pilots on how to handle the aircraft in a real-world simulator as well as in practice. Losing the skill will be disasterous. Also, the pilots should have training in emergency pro
  • Once they have proven themselves to be equal to or superior to human pilots, which would require a proven track record, I see no reason to fear a plane that can land itself. Hell, let's do away with pilots entirely. A plane that is entirely self-piloting is a plane that can't be easily hijacked by anyone with some basic piloting ability, which could solve our TSA problem. Unfortunately, this creates bigger vulnerability to cyber warfare or GPS spoofing attacks.
  • Comment removed based on user account deletion
  • flying is easy. Landing not so much. Especially considering that is is the only mandatory part of flight.

  • Ultimately the question is: is there a better option? If not, then yes of course.

    This is the same answer I currently, inherently provide when I ask myself or am asked about autonomous driving or autonomous anything. While at present autonomous driving is not yet scientifically, statistically safer than human driving (ironically because of the fact there are more humans driving and thus less people following rules, thus computer systems having a harder time reacting to such volatile behavior), I actually alr

  • by BrendaEM ( 871664 )
    There really is no AI.
  • As long as there is a human pilot keeping an eye on it I would not mind it.
    Dont think I would ever set foot in a fully automated plane without a pilot. That would be asking for trouble.

"The vast majority of successful major crimes against property are perpetrated by individuals abusing positions of trust." -- Lawrence Dalzell

Working...