15 ms·

This is a seriously impressive bit of software engineering. It seems as simple as "push a button" to the layman but what happens when: There is a thunders

• michaelmior 7y ago
  It will be insanely impressive if it actually performs well in those scenarios :)
• nabla9 7y ago
  It's to be used only in the event that the pilot is incapacitated. Because it's an emergency feature engaged by a passenger or automatically, it probably can't manage all complex situations. It may be just good weather emergency landing. Check fuel and wind, steer to nearest emergency airport and land. Best effort is better than no effort and dying. > There is a mountain between the normal 3-degree glide-slope and the runway, an "offset approach" I think it can land only to those airports that are it's database with landing approaches programmed it. Cruise missiles have had similar technology for a long time.
  • Daviey 7y ago
    > Cruise missiles have had similar technology for a long time. I try to make my landings a little less eventful.
    • rhombocombus 7y ago
      I prefer to be subsonic before I let the mains touch down (atleast...)
    • AWildC182 7y ago
      A "good" landing is one which people can walk away from and a "great" landing is one where you can use the cruise missile again so by definition, a cruise missile can complete neither a "good" nor a "great" landing by design.
      • OnlineGladiator 7y ago
        Well in theory it could fail in such a miraculous way that everything is intact and it just skidded across the dirt - so while that might be considered a "good" landing it would be considered a "bad" missile launch.
      • serf 7y ago
        there are plenty of loitering munitions that have great landings routinely.
      • exikyut 7y ago
        Then the SpaceX stage1 (or stage2, or however they call it) is a missile that only makes great landings. :D
      • bstrong 7y ago
        "One unusual capability of the Snark missile was its ability to fly away from its launch point for up to 11 hours, and then return for a landing. If its warhead did not detach from its body, then the Snark could be flown repeatedly." https://en.m.wikipedia.org/wiki/SM-62_Snark https://en.m.wikipedia.org/wiki/SM-62_Snark
        • AWildC182 7y ago
          Of course cold war shenanigans would prove me wrong.... I stand corrected
  • mhb 7y ago
    From https://news.ycombinator.com/item?id=21409114 https://news.ycombinator.com/item?id=21409114 During those first few seconds after I hit the Autoland button, the system went through a series of complex calculations and decision-making processes to determine the nearest suitable runway based on runway length, width, and surface; fuel remaining; crosswind component; terrain; obstacles; and general weather information. The system requires an RNAV approach, but beyond that, the runway and weather criteria can be decided by the airframe manufacturer. The system even forecasts its own weather if the nearest suitable runway is a significant distance away, long enough that the current ADS-B or SiriusXM weather may not be valid. It uses the latest weather trend information, for example, to determine if a thunderstorm might move into the runway environment where it intends to land. It will route the airplane around thunderstorms as well as terrain and obstacles, all of which it gets from its internal databases. If en route to a runway it determines, because of changing weather conditions, that another runway is closer or more suitable, it will change its destination. It can even estimate changing barometric conditions and adjust the altimeter—using algorithms. Garmin engineers say the calculated barometric readings are within 0.01 inches of mercury of actual ambient conditions.
  • Xelbair 7y ago
    >> Cruise missiles have had similar technology for a long time. so that's where Boeing got the idea for 737MAX!
  • briandear 7y ago
    Almost airport with runways long enough to land G3000 equipped airplanes will have a published approach. What will be interesting though is how this system handles circle to land approaches.
    • cornellwright 7y ago
      My guess is it would choose a different approach or airport rather than circling if the winds are too unfavorable.
  • zarkov99 7y ago
    Would it be a lot simpler to have a remote pilot access the controls? Why go through all this ?
    • rexaliquid 7y ago
      Establishing, maintaining, and handling the inevitable connection issues of the data connection necessary to remote pilot an aircraft is a lot more work than integrating a "best route" algorithm with existing autopilot features, automated radio communication, and calming cockpit UI.
      • nroets 7y ago
        Until StarLink has proper coverage.
        • denkmoon 7y ago
          Even then, would you really want to rely on that?
      • zarkov99 7y ago
        I was assuming that this was a solved problem, given how prevalent drone warfare is now.
        • ceejayoz 7y ago
          The military can easily afford other aircraft to serve as relays and satellite data connections that a general aviation pilot probably can't. Even then, it's problematic: https://en.wikipedia.org/wiki/Iran%E2%80%93U.S._RQ-170_incident https://en.wikipedia.org/wiki/Iran%E2%80%93U.S._RQ-170_incid...
  • phissk 7y ago
    > Cruise missiles have had similar technology for a long time. Cruise missiles aren't trying to survive the landing.
  • SEJeff 7y ago
    Cruise missiles don't "land" and they don't care about passenger safety :)
  • kevin_thibedeau 7y ago
    > Cruise missiles have had similar technology for a long time. Until GPS guidance came into use they were relatively crude terrain followers requiring a preloaded flight plan over a known terrain profile with sufficient bumpiness to protect against IMU drift. Nothing at all like this technology.
  • iancmceachern 7y ago
    > Cruise missiles have had similar technology for a long time. Yeah but an average human couldn't, and thankfully still cannot buy one. This is feature on a (and more to come) civilian aircraft, two very different things.
• stanski 7y ago
  The marketing material suggests that weather and runway conditions are taken into account when selecting a landing site. So I would assume the answer to all of those is that it won't pick that runway. And if the aircraft is low on fuel or isn't making enough power to make that runway... Well, the usual happens - a forced landing, aka a crash landing. I don't think this is meant to be used as a cop-out by the pilot. "My engine quit! You deal with this, computer!"
  • cjrp 7y ago
    In the Vision Jet I believe low fuel / no suitable airport in range leads to automatic CAPS deployment by the autoland system
• throw0101a 7y ago
  Thunderstorm: it can use weather broadcast over ADS-In UAT. Mentioned in the AVweb report. Closure: certainly a problem. Mountain: can use worldwide topographic information in its database to route around. See AVweb Crosswind: mentions getting local weather. Engine: Autland cannot help. But on the Cirrus you'd pull the CAPS cord and float down.
  • pc86 7y ago
    CAPS is not supposed to be used because you have an engine out under most circumstances - you are perfectly capable of dead-sticking a Cirrus onto a runway just like any other small GA plane. And it's not a gentle float, it's ~1700fpm or about 20mph, straight down. The airframe and seats absorb a chunk of it but compared to a 500-700fpm descent on an engine-out landing, pulling CAPS because your engine dies with any sufficient landing area in range (including open fields) will rightfully earn you quite a bit of ridicule.
    • guidoism 7y ago
      The insurance companies apparently want pilots to pull the CAPS rather than try to dead stick it. That’s what I was told when I got checked out in the SR22 by the flight school. They drilled it into me. What happens when blah blah blah? Pull the CAPS. That was the answer to everything.
      • Scoundreller 7y ago
        CAPS sounds great over open water or desert, but not so great over a populated area. At least a non-pilot co-passenger might want to try their luck gliding to open area and then pulling it.
        • DuskStar 7y ago
          A 20mph vertical impact into someone's roof really isn't the worst of outcomes.
• joshvm 7y ago
  > During all phases of flight it communicates with passengers and appropriate air traffic control facilities regarding the new flight plan route and estimated time until landing. Which would suggest that if you need to be vectored to a runway, the aircraft can negotiate that with ATC automatically (and perhaps ATC can advise which is the most appropriate diversion?) This in the situation where the pilot is incapacitated, but the aircraft isn't in an immediate risk of crashing (e.g. you're cruising, but there's bad weather immediately below you). It should also be able to use stuff like D-ATIS to get airport conditions. That seems optimistic though, perhaps it just announces its intentions and hopes ATC clears a path?
  • kejaed 7y ago
    Yes it’ll squawk emergency transponder and broadcast on the radio.
  • ceejayoz 7y ago
    > Which would suggest that if you need to be vectored to a runway, the aircraft can negotiate that with ATC automatically (and perhaps ATC can advise which is the most appropriate diversion?) I strongly suspect it just announces itself. "N123FOO DECLARING EMERGENCY; PILOT INCAPACITATED, AUTOLAND ENGAGED; CLEAR RUNWAY 21L; 12 MILES OUT" sort of stuff.
    • usrusr 7y ago
      Still makes me wonder: do you really need/want all that automation or would it make more sense to prepare for putting a remote pilot into the loop? Not for virtually holding the stick, but for looking at telemetry, communicating and doing all the high level decisions.
      • eternalny1 7y ago
        This is what some of the new "fully automated" semi-truck companies are doing. Remote truckers in an office, similar to a UAV drone pilot.
• dsfyu404ed 7y ago
  >but what happens when... You handle those things manually by setting proper inputs to the normal non-emergency autopilot or flying without autopilot like you would in any other aircraft equipped with an autopilot that is capable of landing. I assume "one off" approaches are a non issue because they are using a database of supported airports with the necessary approach details since they change slowly and the list is short enough. That's traditionally how this kind of thing is handled. This is not a replacement for a pilot and is not portrayed as such.
• Stevvo 7y ago
  In the case of the mountain, the avionics have a terrain database built in, so it can follow an LPV WAAS approach from a fix above the terrain. I don't know if this can deal with the engine out, but in theory there isn't any reason it couldn't. There is an IPad app called "Xavion". With an engine out, Xavion will calculate an approach for you follow to a suitable airport, based on the aircraft's best glide speed + required runway length. Interestingly it doesn't use the published best glide speed and stopping distances, the app uses your flight history to figure out the aircraft's actual best glide and stopping distance.
  • o-__-o 7y ago
    Head and tailwinds are real
• asdfman123 7y ago
  Question: how close are we to totally self-flying planes? I feel like we could easily get there before we got self-driving cars, if we wanted to. Would the costs for short hops be comparable to a commercial airplane ticket? Because it would be awesome to charter small flights from an app to jump to a nearby city.
  • Stevvo 7y ago
    The technology already exists, it's a much easier problem than self driving cars. Spacecraft can fly themselves to other planets and return. That is a massively more complex problem. It's a problem of trust, and regulation, not a technical one.
  • ocdtrekkie 7y ago
    There's been discussions for years on whether or not the level of automation in modern aircraft causes pilot training to lapse due to misuse. Self-driving planes are much easier than self-driving cars because there is empty space on all sides of the plane pretty much at all times. Autopilot is already generally speaking "a thing", with the exception of takeoff and landing. Bear in mind even in takeoff and landing scenarios, unlike a public roadway, the airplane has pretty exclusive use of the runway as well. Pilots are very unlikely the primary cost of aircraft trips. Things like jet fuel are probably much more of the cost driver.
  • forgottenpass 7y ago
    >I feel like we could easily get there before we got self-driving cars, if we wanted to. The environment has shown quite clearly that we don't want to. We'd probably already have it if we did. The safety design of aviation relies on having multiple highly competent people in the loop on every flight and exercising their judgement on dozens of things. And that's for the best-case scenario, they also have to be capable of choosing and executing fallback plans when part of the overall system (the plane, air traffic, radio navigation, flying conditions, ground conditions, etc...) degrades or enters an unexpected state.
  • tim333 7y ago
    Engine failure on takeoff is tricky to deal with. See Sullenberger landing on the Hudson for example https://en.wikipedia.org/wiki/US_Airways_Flight_1549 https://en.wikipedia.org/wiki/US_Airways_Flight_1549
• griffinkelly 7y ago
  I've flown airplanes with solid GA plane autopilot that'll fly down to decision height, but you still have to put in flaps, gear, change throttle. To have an airplane do all that on its own is quite impressive, particularly to do a 'nearest' type input on a Garmin then to land it on the appropriate runway, still confuses me sometimes if its a new airport I've never been at. I'd be interested to know if it declares an emergency, squawks 7700 and controllers are expected to give it a priority and right of way?
  • o-__-o 7y ago
    Why? Airbus and Boeing and even smaller planes like bombedaire have had this capability. With ads-b requirements on all planes it puts GA aircraft in line with jets (ads-b requires transmitting internal information that autopilot also processes) What I’m trying to say is that the auto land functionality is not a feat, it’s like self parking cars.. been around for 20 years just now cost efficient and safe enough for the masses to use. In a boeing 787, the pilot still needs to squawk 7777 and they should try comms, don’t know why a GA plane would be different..
    • outworlder 7y ago
      > the pilot still needs to squawk 7777 and they should try comms, don’t know why a GA plane would be different.. The system being discussed is for when you no longer have a pilot.
      • crooked-v 7y ago
        Yes, this is not routine autopilot landing, this is "the pilot just had a massive heart attack and the only other people in the plane are a 12-year-old and an armless nun" landing.
    • andrewg 7y ago
      Autoland, as used in a capable airliner, is much more limited than this. For an airliner, it's a procedure with lots of configuration and monitoring done by a pilot, in cooperation with air traffic control, which usually concludes with descent on a 3-degree or so path through touchdown and braking. Not to mention things like flaps and landing gear are still usually manually selected. This is much more impressive, not the least of which because the system itself figures out the best place to land (taking into account operational constraints), how to get there (taking into account weather avoidance), and how to be at the right speed, altitude, configuration, etc. to perform the final approach to landing. So to use your analogy, getting a bit closer to a self-driving car than just self parking!
• skizm 7y ago
  "Conditions are not suitable for autoland functionality. Please navigate to a location with a clear view of a runway and try again. Sorry for any inconvenience."
• lisper 7y ago
  From https://www.piper.com/model/m600sls/ https://www.piper.com/model/m600sls/ "The routing algorithm determines a path to the Final Approach Fix (FAF) of the chosen published approach that avoids terrain, obstacles, and significant weather." That handles all of your contingencies except an engine failure. If you have an engine failure and a pilot who can't dead-stick a landing you're probably screwed. But you're still no worse off in that case than you would have been without autoland.
  • jacquesm 7y ago
    I disagree with that. You might now be a pilot who has lost his experience in how to land an airplane and so will do a much worse job of it compounded by the emergency situation. Every landing is also training.
    • lisper 7y ago
      Yes, that true, but this autoland system is not intended for routine use. It's for emergencies only.
      • jacquesm 7y ago
        Ah, 'you' being 'the passenger and/or the incapacitated pilot' rather than the regular and healthy variety of pilot. Ok, now I understand your comment. Thank you for clarifying.
  • appstorelottery 7y ago
    I think this a case for better GA trining. Expecting engine failure on takeoff, building the counterintuitive pitch-down reflex to avoid stall, and so on. Knowing your minimum maneuverable speed (and marking it on your air speed indicator). Expecting something to go wrong is the key mindset. A GA death happens roughly once a week for the reasons you outline. Emergency situations are not time to learn new skills, you need to have it down.
    • Zak 7y ago
• savrajsingh 7y ago
  I’d hope that autoland connects to Garmin HQ and then they can coordinate and also just call up the tower and tell them what’s going on. Managing airspace and radio communications is the tricky part imo
  • kawfey 7y ago
    Uncontrolled airspace is much more of a concern, since there is no tower HQ can call.
  • neom 7y ago
    I'd presume the plane will also start to squawk 7700, shouldn't be an issue then.
  • FabHK 7y ago
    Autoland will squawk the emergency code 7700 on the transponder, and announce itself on the emergency frequency 121.5 and the local CTAF (common traffic advisory frequency, IIRC, which coincides with tower frequency if the airport is towered).
• __m 7y ago
  Well the plane crashes i suppose
• rdevsrex 7y ago
  No doubt this is real engineering. I'm sure they had to do a lot of software verification work.
• rubicon33 7y ago
  It's just shortest path traversal no? Represent each airport as a node in a graph with all relevant data (distance, weather, etc.) ... Traverse the graph and find the shortest path. https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm Easier said than done I'm sure ;)
  • rubicon33 7y ago
    Confused as to why this is being downvoted. If there's a better way to implement it, by all means, lets discuss that! It's an interesting tech challenge... why just downvote?
    • tonyarkles 7y ago
      I agree with you! The word "just" in their post may be prickly. Also, it's not really a graph search as stated, just a "choose the airport with the lowest weight". That, in theory, chooses the airport to land at, but not... the rest of it. Yes, choosing a route would be essentially a graph search, with weather at intermediate nodes adding to edge weights, etc, but even after you've computed a target path, you're going to be continually refining that path (e.g. the thunderstorm moves in a way you didn't predict). And... flying the aircraft itself. Autopilot in straight & level flight isn't too big of a deal. Autopilot tracking a path is doable (including ascent/decent). Auto-throttle is common on larger airliners but not in GA aircraft. Automatically configuring landing gear and flaps, as appropriate for weather conditions, is (afaik) not done automatically even in commercial airliners. And doing all of this, reliably, with the assumption that the pilot is incapacitated... that's friggin' cool!
• saganus 7y ago
  What kind tech stack/development processes do you think are needed for something like this? At a very naive level, for example: typed language vs dynamic; formal verification of the code or not (or to what extent); what kind of QA? (can't even imagine how to test this "in production" given the amount of edge cases like the ones you describe; hw and sw redundancy; to name a few. I have zero experience with airplane software but I find it fascinating considering how safe flying is.
  • HeyLaughingBoy 7y ago
    I'd imagine that it begins well before that with Requirements Management: probably more important than any other single thing since that drives the entire development process. Anyway, I'm sure there used to be a blog for the Dreamliner test team, but I can't find it now, so here's this instead: https://www.wired.com/2010/03/a-look-inside-the-brains-of-boeing-flight-test/ https://www.wired.com/2010/03/a-look-inside-the-brains-of-bo... And specifically for code, look here: https://ldra.com/aerospace-defence/standards/misra-cc/ https://ldra.com/aerospace-defence/standards/misra-cc/ and (JSF, not commercial) https://www.f35.com/about/life-cycle/software https://www.f35.com/about/life-cycle/software
  • outworlder 7y ago
    > typed language vs dynamic You are not approaching this from the correct angle. "dynamic" and "static" types is a red-herring for safety. Yes, when we are rushing some feature in our day jobs we can try to call functions with their wrong types. But these are not critical systems. If you DO have a critical system, you are going to have to invest _massive_ engineering resources. You are going to exercise all code paths. If you get a "type" error, you haven't done your job correctly. Because this was a code path not exercised before, and that should not happen. The additional type checks at compile time are a drop in the bucket compared to all the testing you have to do. That said, most such applications will be very traditional. C or C++, maybe ADA. Despite having types, the C family you give you a whole lot more failure scenarios. For instance, it's very easy to corrupt your entire process memory in C, that is something that not many "dynamic" languages will let you do. And yet people build mission critical systems with them. That's not the only possible answer though. For an interesting autonomous system, please look into NASA's Deep Space 1, in particular the Remote Agent. Written in Lisp, it had spacecraft controls for a few days. It could self-diagnose and correct issues. For a long time, it was the most advanced piece of software to ever control a spacecraft. And Lisp is "dynamic" typed (but strongly typed) Now, they do test in "production", but it's more like they have lots of staging environments. Some of them may be simulators, some of them may be replicas. At the end of the day, this is no different from any other software: it has inputs (airspeed, altitude, heading, roll, pitch, yaw, servo positions, charts, weather radar, etc) and outputs (controlling the servos, changing radio frequencies, etc). You can simulate some or all of them.