Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
×
Transportation AI

Airbus Close To Landing Fully Automated Passenger Jets (theregister.com) 76

UpNext, Airbus's future technology-focused subsidiary, reported on Thursday that it has entered the final three months of testing tech it hopes will automate the process of getting a plane from the air to the gate. The Register reports: The system, called DragonFly, tackles automated operations like diversions, landings, and taxi procedures through a combination of sensors, computer vision algorithms and robust guidance calculations. Airbus pitches the system as an extra layer of safety for emergency operations. "In the unlikely situation where a crew is unable to control the aircraft, DragonFly can redirect the flight to the nearest appropriate airport and facilitate a safe landing," enthused the aerospace corporation. The eventual hope is for the technologies to pave the way for automated landing -- or at least compensate for a less than perfect pilot during an emergency situation. If the captain had the fish, for example.

A marketing video of the Airbus UpNext DragonFly details that a safe landing feature is included in the Automated Emergency Ops application. It works by detecting the most suitable airport for landing and calculates a trajectory to get there, with consideration for weather, military zones and other factors. Airbus assures viewers that Air Traffic Control (ATC) and Operations Control Center (OCC) communication links are in place. The video does not, however, explain how the aircraft communicates with air traffic control for clearance into controlled airspace if the pilot is incapacitated, as that task is completed through human verbal interaction. An automated landing is assisted by sensors that enhance the view of the runway, computer vision algorithms, and guidance computation.

Furthermore, the demonstrator is kitted out with a pilot taxi assistance application to manage its maneuvers on the ground in a heavily trafficked airport. Air traffic control clearance is interpreted and translated into taxi guidance cues. Crew receive audio alerts in reaction to obstacles, assisted speed control and an interactive airport map. The taxi assistance element was tested at Toulouse-Blagnac Airport. The subsidiary reckons one day DragonFly will allow for automatic landing at any airport, regardless of whether the ground equipment is equipped for such landings.

This discussion has been archived. No new comments can be posted.

Airbus Close To Landing Fully Automated Passenger Jets

Comments Filter:
  • and do an Boeing 737 Max 8 with low # of sensors?

    They need to have at least 3 sensors and for something maybe 5 and have IF X fail then land at nearest airport right now.

    • by shanen ( 462549 )

      Rather a weak FP, but maybe it was the time pressure. You are touching on deep waters, but with a stick rather than a sponge?

      The fundamental philosophical difference between Boeing and Airbus is quite important on this topic, but given today's time constraint I'll just reference The Glass Cage by Nicholas Carr (and check the discussion to confirm that there is no other reference). Someone on Slashdot must have read it?

  • by CaptQuark ( 2706165 ) on Saturday January 14, 2023 @02:19AM (#63207672)

    I seem to remember aircraft have had a form of autonomous/automated landing since the 1980s. It is called Autoland. https://en.wikipedia.org/wiki/... [wikipedia.org]

    • by sxpert ( 139117 )

      this appears to me much more automated.

      • by Freischutz ( 4776131 ) on Saturday January 14, 2023 @03:23AM (#63207716)

        this appears to me much more automated.

        Yes, airliners can't currently land 'automatically' without ground based aids. This project presumably aims at being able to land with nothing but an AI at the helm and no external aids at all. Whether AI landings will also turn out to be possible in an emergency given any combination of unanticipated systems failures and external events under any weather conditions whatsoever at any airport you care to name that has a suitably long runway is something I'm going to reserve the right to doubt. There are some seriously tricky runway approaches out there at runways you'd normally not even dream of trying to land an airliner at but that you might have land at anyway because the alternative is the 100% certainty of crashing with the loss of all lives on board. I'd regard this more as another step on the long road towards eliminating pilots completely than it is an immediate threat to the job security of every pilot on earth. In the mean time this is going to be a backup system that's better than current ones, which will enable you to land at, say JFK with both pilots dead. Trying to replicate that at Thule: https://youtu.be/ngaHw8P1RMM [youtu.be] ... might be harder and Thule isn't even the worst place you can end up diverting to.

        • There are some seriously tricky runway approaches out there at runways you'd normally not even dream of trying to land an airliner

          Your logic seems backward to me.

          Landing at an unfamiliar and "tricky" airport during an emergency, perhaps low on fuel or with a damaged engine, so no possibility of a "go-around", is exactly where an AI is needed most.

          Unlike a human pilot, an AI can pull the airport info from a database and land as well on its first attempt as a human would on his 20th.

        • I'd regard this more as another step on the long road towards eliminating pilots completely than it is an immediate threat to the job security of every pilot on earth.

          I'd guess the nearest-term practical effect would be raising the size of the plane / number of passengers that is allowed to be piloted by an individual without a human copilot onboard.

        • Your comment is quite accurate but I'm not sure how it fits with the goals stated in the fine summary. If the crew is incapacitated (which seems the be the target situation and quite rare), the ability for the plane to get itself down is quite remarkable. And it doesn't have to be even at the target airport or necessarily in bad weather. If the crew declares itself incapacitated and the plane can take itself to the nearest airport with good weather and a long runway, that's fantastic.

          Now I admit that th

          • this is really a sneaky way to try to get to single-pilot flying.

            Everything I'm hearing about the push for single-pilot flying leads me to believe the same. I'm very much against that. Aviation safety depends on redundancy of every critical system, and the pilot is, of all systems, BY FAR the most critical.

            Having said that, I'm very much in favor of technology to enhance safety when Bad Things happen, because, unfortunately, sometimes they do. One pilot eats the steak, the other one chicken, but unb

    • by fahrbot-bot ( 874524 ) on Saturday January 14, 2023 @03:55AM (#63207732)

      It's called Gravity. Works every time, all the time.

      • Hell yeah. With the help of gravity I'm also pretty close to landing passenger planes.

      • It's called Gravity. Works every time, all the time.

        Not if you are going fast enough.

      • As Launchpad McQuack said, any crash you can walk away from is a good crash!
        • I was surprised to learn that, largely because of truly outstanding and heroic actions on the part of pilots and cabin crew, people do survive many kinds of crashes. Not usually the ones where you slam into the ground at a 90 degree angle at 600 knots. But quite a few of the others. Often, injuries and deaths result not from the impact itself, but from the ensuing fire, which is part of why fuel is dumped and cabin crew are always prepared to evacuate quickly.
    • I seem to remember aircraft have had a form of autonomous/automated landing since the 1980s. It is called Autoland. https://en.wikipedia.org/wiki/... [wikipedia.org]

      Yep, but this new one can taxi to the gate.

      Landing an aircraft automatically isn't difficult. It's a straight vector to the runway and there's no obstacles to avoid.

      Like most things "aircraft" though, people freak out about it.

      (The same people who drove to the airport in a tin box while negotiating all sorts of obstacles)

      • by dryeo ( 100693 )

        Airports in mountainous valleys can be tricky, hard turns as descending to the runway kind of approaches along with mountain weather, up and down drafts.
        Example, https://en.wikipedia.org/wiki/... [wikipedia.org]
        To quote,

        Due to the mountainous terrain impinging on both runway approaches, there is no possibility of a straight-in approach. The airport is therefore certified for day operations only, and the glideslope on approach is set to a steep 5.0 rather than the standard 3.0. The instrument approaches to Castlegar are con

      • Instrument landings in good conditions are fairly routine, but conditions aren't always good, you can encounter unexpected problems, and sometimes the best you can hope for is a landing that people can walk away from.

        You really need pilots. (Plural. More than one.) Not for the 98% of the time when things go right, but for the 2% of the time whey they don't.

    • by hackertourist ( 2202674 ) on Saturday January 14, 2023 @06:57AM (#63207936)

      Autoland requires the pilots to navigate the aircraft to a fixed starting point where the aircraft can pick up radio beacons aligned with the runway (ILS or MLS). The autoland system can then put the aircraft on the runway, and brake to a halt. Several categories of autoland exist, the best ones are certified for use in 0 visibility.

      This Airbus system extends that on both ends: it can start at cruising altitude, find the nearest airport, land the aircraft, and then taxi off the runway to a gate.

      • Taxing to the gate seems like a pretty useless feature in an emergency. But the ability to find the nearest airport and get on the ground seems quite useful in the exceptionally rare case that something happened to both of the pilots.
        • Getting the aircraft off the runway is a very useful feature. In an emergency, you'd have to drive a staircase to the aircraft plus a relief pilot, and all the while that aircraft is sitting the middle of a runway and blocking your entire airport.
          In a non-emergency, think of it as extending your autoland. The aircraft will be able to find the gate it's supposed to go to even if the pilots can't see shit.

          • Which is about seven orders of magnitude better than having the aircraft crash and eight orders of magnitude better than having it crash at your airport! So in relative terms, I think it's still quite small!
        • You want to be off the runway, be able to evacuate quickly in case of fire or people needing medical attention, and preferably someplace adequately lit or at least clear of dark roadways. (People have survived plane crashes, then died after being run over by emergency vehicles that could not see them.)
          • I think I should have phrased that differently. You are right that being able to get off the runway is good *incremental* value on top of actually being able to land. But making the landing is probably 80-90% of the value. However, in an emergency, you don't taxi to the gate at all. I've actually been (as a passenger) on planes that have made real emergency landings. The airport had us get off the runway and head to a remote parking area which was more accessible to fire crews than the terminal. Plus t
    • And even older than that was the L-1011 system which never "took off". https://www.lockheedmartin.com... [lockheedmartin.com] May 25th, 1972.
    • Lockheed L-1011s could take off and land automatically back in the 70's.
    • by Jemm ( 747958 )
      The push is to reduce pilots in the cockpit to one person from two.
      • Which no one wants except the airline companies, and which will likely reverse 50 or more years worth of improvements in commercial aviation safety. :(
    • by tri44id ( 576891 )

      Private planes with advanced instrumentation from Garmin have had emergency autoland capability for several years now. (https://www.getinpulse.com/how-garmin-autoland-works [getinpulse.com])

      It can go as far as automatically determining when the pilot is incapacitated, notifying air traffic control of an emergency, figuring out the nearest airport within 200 miles, navigating there, and landing itself. Currently certified on planes from five different manufacturers, in the US and Europe.

      Of course, doing the same thing wi

  • by robbak ( 775424 ) on Saturday January 14, 2023 @02:46AM (#63207686) Homepage

    If - big if - they can prove that they can get a plane safely on the ground without a pilot, then the automation becomes the backup to a single pilot, and having two pilots becomes unnecessary. In their opinion.

    I feel that a single person cockpit would be a very lonely place to work.

    • by Roger W Moore ( 538166 ) on Saturday January 14, 2023 @05:17AM (#63207804) Journal

      then the automation becomes the backup to a single pilot

      Only for physical failure modes of the one pilot. If the sole pilot fails mentally then the plane is doomed since, unlike a human co-pilot, automation can always be overridden and probably can't even detect mental issues. Plus being alone is only going to exacerbate mental issues further.

      • then the automation becomes the backup to a single pilot

        Only for physical failure modes of the one pilot. If the sole pilot fails mentally then the plane is doomed since, unlike a human co-pilot, automation can always be overridden and probably can't even detect mental issues. Plus being alone is only going to exacerbate mental issues further.

        100% agreed.

        Plus, I'm not sure it would even save that much, it looks like the co-pilot only earns ~$100/hr, a fraction of the main pilot [epicflightacademy.com]. So that's maybe $1/seat? Not really worth the safety risk and public blow back.

    • If - big if - they can prove that they can get a plane safely on the ground without a pilot.

      They've been doing it since the 1980s, so...

      (and it's waaay easier than, eg. self-driving cars)

    • Why even have a pilot? Just install a second computer.
    • Have the plane flown remotely from the ATC. Solves a lot of issues.
    • Aircraft can already land on their own - If you have seen any of the "can an untrained person land an airliner" then all the ones who succeed just program the autopilot to land the plane, and apply the breaks at the end ...

      The reason they don't normally is you want alert pilots who can detect issues and land safety regardless - if they just always let the autopilot land and fly it then they would have no experience ..

  • When Boeing does this, there'll be no mention in any manuals or training films. They'll keep it a surprise.
  • by Todd Knarr ( 15451 ) on Saturday January 14, 2023 @03:08AM (#63207706) Homepage

    I suspect, if it's for emergency operations, it handles clearance into controlled airspace by a signal that tells ATC "I'm an aircraft under automated control. I'm experiencing an emergency and require expedited clearance and ATC priority assistance. Please confirm.". ATC would respond with a signal (possibly via voice, with the aircraft doing basic voice recognition) saying either "You are cleared direct to any runway." or "Enter emergency holding pattern and await clearance.". You can simplify a lot just by treating it as an emergency and getting everything else out of the way.

    • I think an emergency at that level would always result in getting the expedited clearance. When the Concorde flight caught on fire, if I remember correctly, ATC had the whole airspace cleared in thirty seconds or less which is astonishing but that's why they are the experts in what they do.
  • by cowdung ( 702933 ) on Saturday January 14, 2023 @03:20AM (#63207712)

    Knowing Airbus.. they'll give each pilot an Autoland feature.

    PIlot 1 will ask to land in Airport A.

    Pilot 2 will ask to land in Airport B.

    So the Autoland will take the average of both commands and land there.

    • Re:Average of both (Score:5, Interesting)

      by ShanghaiBill ( 739463 ) on Saturday January 14, 2023 @06:07AM (#63207892)

      You appear to be joking, but this is similar to what happened to Air France 447 [wikipedia.org].

      Sensors failed, the plane began to stall, and the cockpit crew was confused. The captain moved his stick forward to push the nose of the plane down, gain speed, and recover from the stall (the correct response). Meanwhile, the co-pilot pulled back on his stick for unknown reasons.

      The Airbus 330 AVERAGED the two inputs, canceling them out. Neither pilot was paying close attention to what the other was doing, and they were confused as to why the plane was unresponsive to their input. The plane stalled, fell from the sky, and 228 people died.

      This would not have happened on a Boeing. Boeing allows input from only one stick at a time. If the captain was flying, the plane would have been fine. If the copilot was flying, the captain would have immediately seen he was responding incorrectly and taken control.

      • by ghoul ( 157158 )
        Boeing would have just have a piece of software that ignores both pilots and put the plane into an unrecoverable dive. "Boeing trusts pilots. Airbus trusts software is just a meme"
      • by GFS666 ( 6452674 )

        You appear to be joking, but this is similar to what happened to Air France 447 [wikipedia.org].

        Sensors failed, the plane began to stall, and the cockpit crew was confused. The captain moved his stick forward to push the nose of the plane down, gain speed, and recover from the stall (the correct response). Meanwhile, the co-pilot pulled back on his stick for unknown reasons.

        The Airbus 330 AVERAGED the two inputs, canceling them out. Neither pilot was paying close attention to what the other was doing, and they were confused as to why the plane was unresponsive to their input. The plane stalled, fell from the sky, and 228 people died.

        This would not have happened on a Boeing. Boeing allows input from only one stick at a time. If the captain was flying, the plane would have been fine. If the copilot was flying, the captain would have immediately seen he was responding incorrectly and taken control.

        A little clarification on your excellent response if I may. The fundamental problem with Airbus's cockpit is that it has side sticks for control. These side sticks are electronic force sensors with a handle so 1) you can't see them physically move and 2) as you correctly point out, the flight control computers can have two totally different inputs into them. Boeing, to their credit, does not do this. They have a physical old style control wheel on a control column that MOVES and is MECHANICALLY connected to

      • Re:Average of both (Score:5, Informative)

        by Cochonou ( 576531 ) on Saturday January 14, 2023 @12:56PM (#63208454) Homepage
        Not really. Please read again the crash report [bea.aero].
        - The captain was on the back seat and was never in control
        - When both pilots use inputs, there is a loud "dual input" audio signal in the cockpit
        - The plane had already stalled when dual inputs occurred
        - Boeing airplanes do not have sticks, they have yokes. They are mechanically coupled.

        The plane crashed because the pitot tubes froze, because the pilots did not recognised the need to apply the unreliable speed procedure, because they were not trained in high altitude stalling (the stall alarm sounded 75 times, but nobody acknowledged it, even the captain), and because the crew resource management was abyssal. A sad event.
  • by Anonymous Coward

    It's been 2 decades now that ATC Centres and many International aircraft have been communicating by text. Beyond that, automatically changing the transponder to 7700 will be sufficient to have everything in its way moved and priority for landing ensured.

    One might assume that is trivial to include such notifications in a licensing regime that would surely be required by the complexity that such a system invites.
    If it can fly the plane it can probably send a couple of text messages.

    • by cstacy ( 534252 )

      It's been 2 decades now that ATC Centres and many International aircraft have been communicating by text. Beyond that, automatically changing the transponder to 7700 will be sufficient to have everything in its way moved and priority for landing ensured.

      One might assume that is trivial to include such notifications in a licensing regime that would surely be required by the complexity that such a system invites.
      If it can fly the plane it can probably send a couple of text messages.

      Maybe eventually there will be a new "77" code. There's already "emergency" and "lost comm" codes. Assigning an "emergency autoland" code would be easier than actually implementing any automation to request and receive clearances. It'd just be the same as with 7700, which like you say, gets everybody out of the way and you're cleared to do whatever you want. The airspace and airport is essentially shut down except for you. A new code would be informational.

      (Otherwise, what are we talking about with clearan

  • by buck-yar ( 164658 ) on Saturday January 14, 2023 @05:19AM (#63207810)

    How will they keep bad guys from taking over this system?
    Human pilots can
    *make mistakes
    *get depressed and intentionally crash the plane
    But at least they can't be taken over remotely and crashed or held for ransomed (assuming this didn't happen to MH370)

    • How will they keep bad guys from taking over this system?

      I assume the pilot will have a 'cancel' button if he sees all the screens turn red and hears a voice warning about the aircraft "Entering emergency mode".

      Besides, what's the worst that could happen? The aircraft might land earlier then expected at a different airport?

      • The worst case scenario is much worse than that. If the plane were to land early at an unexpected airport, it would be considered an emergency and ATC would respond appropriately. ATC does their job phenomenally well. But nobody is perfect. In the scramble to deal with the emergency they could make a mistake that has safety consequences. Part of a system like this would require some mechanism to deal with improper activation. It might require say the flight attendants to unanimously declare the pilots
  • by cstacy ( 534252 ) on Saturday January 14, 2023 @05:28AM (#63207826)

    I am not enamored of the idea of any "close landings" by Airbus or anyone else. It sounds like "In the event of a water landing..." except on the ground.

    Seriously though, this is an expected evolution of the traditional ILS-based autoland. But general aviation airplanes ("little airplanes" with four seats) already have a GPS-based autoland capability. For example, the one called "SafeHome". The pilot is incapacitated, so the passenger pushes literally a Big Red Button, and the airplane finds a suitable airport and lands. It doesn't need ILS or anything. (I haven't read the manual, so I don't know if it changes the transponder code, but it's not calling ATC or anything.) After it lands, someone has to manually taxi. This has been a feature in these planes for a few years.

  • Scarebus. Sometimes automation goes wrong, when the plane thought it was landing [youtu.be].
  • You put in an autopilot with one backup human pilot on board who does no handflying 95% of the time and then you expect him to takeover in an emergency and save the plane? Not gonna happen. Keep the backup pilot on the ground. He will get regular practice hand flying remotely planes whose autopilot get into trouble so he will be in practice. For cases where the remote link gets disabled use whole plane parachutes.
  • ...not the Boeing way of 'landing'.

  • Automated taxi (Score:4, Interesting)

    by WaffleMonster ( 969671 ) on Saturday January 14, 2023 @09:45AM (#63208118)

    Only reason they would bother with engineering required to make ground taxi work is if they are intending to replace pilot(s). They just don't want to say this is their goal for PR reasons.

    • Thank you. I stated the same thing in a reply to an earlier poster. But your post was chronologically sooner than mine. This certainly seems like it's laying the groundwork (bad pun not intended but I like it) for single-pilot operations.
  • by AlanObject ( 3603453 ) on Saturday January 14, 2023 @11:31AM (#63208302)

    Although you can automate a lot if not all of a pilot's tasks, doing so tends to put pilots in the air that are not really good at doing the tasks that the automation took over. Training pilots is, after all, very expensive.

    A few years back there was a botched landing at SFO. From the report that I got through the lay press it appeared that the pilot didn't seem to have a good grasp of the stall speed of the aircraft. The airliner stalled in short of the runway. So we may be there already.

    But we can look forward to incidents where automation failed and the "backup" pilot wasn't able to cope with what was an avoidable accident.

  • That should cut down on loading and unloading times since the passengers are now automated

  • Then, I will believe all these other developments. Really, we build machines that cost double and triple digit millions but we still don't have planes that just don't crash. I don't understand why plane software cannot basically prevent crashes. It can't be that hard to just code the 'do not allow the plane to hit the ground unless it's doing so in controlled flight at an airport on the runway under specific parameters'. They can use as many sensors as possible and have as much redundancy as possible becaus
    • Crash risk is trivial but fearful simples are phobic about air travel yet drive to work every day.

      Any system can fail and the risk of air crashes by civilization-based airlines is essentially nil. The point of diminishing returns was likely years ago. Life is not infinitely valuable, humans merely fear losing it.

    • Please study some actual crash and serious incident reports.

      Planes almost never crash unless MULTIPLE things go wrong simultaneously. Usually involving things like human error, bad conditions on the ground, fan blades that slice through all three hydraulic systems at once, and the like.

      Planes themselves are remarkably safe, and commercial aviation in general is remarkably safe.

      And for every crash caused by human error, dozens if not hundreds are prevented by pilots thinking their way out of really, really

  • The Lockheed L-1011 "TriStar" was the first commercial airliner delivered from the factory with fully autonomous flight capabilities from launch to landing, an astounding achievement pre-GPS, largely enabled by an advanced military-grade inertial navigation system.

    FMI: https://en.wikipedia.org/wiki/Lockheed_L-1011_TriStar

  • I know how well the auto scan machine works at the supermarket.

Success is something I will dress for when I get there, and not until.

Working...