Next-Gen Autopilot Puts A Robot At The Controls

While the concept of automotive “autopilots” are still in their infancy, pretty much any aircraft larger than an ultralight will have some mechanism to at least hold a fixed course and altitude. Typically the autopilot system is built into the airplane’s controls, but this new system replaces the pilot themselves in a manner reminiscent of the movie Airplane.

The robot pilot, known as PIBOT, uses both AI and robotics technology to fly the airplane without altering the aircraft. Unlike a normal autopilot system, this one can be fed the aircraft’s manuals in natural language, understand them, and use that information to fly the airplane. That includes operating any of the aircraft’s cockpit controls, not just the control column and pedal assembly. Supposedly, the autopilot can handle everything from takeoff to landing, and operate capably during heavy turbulence.

The Korea Advanced Institute of Science and Technology (KAIST) research team that built the machine hopes that it will pave the way for more advanced autopilot systems, and although this one has only been tested in simulators so far it shows enormous promise, and even has certain capabilities that go far beyond human pilots’ abilities including the ability to remember a much wider variety of charts. The team also hopes to eventually migrate the technology to the land, especially military vehicles, although we’ve seen how challenging that can be already.

19 thoughts on “Next-Gen Autopilot Puts A Robot At The Controls

  1. I’d like to see him do a full preflight of the aircraft, then climb into the cockpit without assistance like the rest of us have to do. And nobody mentioned anything about making him earn the right to solo by logging hours with an instructor before being turned loose on the world.

    1. i’m skeptical of its capabilities but fwiw the article did mention earning the right to solo by logging hours with an instructor. that’s what they’re doing in their simulators and testing program, which is mentioned in the article. it’s a different testing procedure than human pilots but it doesn’t sound any less rigorous. the neat part of automation is that you only do that once and then clone the certified pilot…but you definitely do put it through trials before letting it out.

      1. Do be fair, autopilots are for those boringly regular situations where humans do horribly. In emergencies, humans “tend” to do better since they can handle “out of bounds” errors more gracefully. Not always, but often enough.

        Anyone who writes code knows that 10% of the code (if not less!) does the core logic, it’s all the error and exception handling that adds up and makes things difficult. And of course us humans have varying levels of ability to handle exceptions as well.

  2. It would be interesting to see cockpit footage of the training and certification sessions. Do emergency maneuvers account for loads and stresses on the airframe and occupants?

    Which aircraft under which conditions is this certified for? Is there an LLM AI giving air traffic control plausible, if not entirely accurate answers?

    And as for edge cases framed as the trolley car problem, let’s not go there.

  3. I find systems like this a little funny. So you have a robot which has a computer for a “brain”, as an example the autopilot/robot/computer might decide to preform some input, like pitching up. To accomplish this requires a rube goldbergian series of steps. to pitch up the flight stick or yoke must be pulled back. so because the robot is humanoid android it needs to preform a set of inverse mechanics calculations to figure out how to drive each of the motors or servos to produce that action. Then any number of motors need to move in unison to actually create that motion. which has the effect of moving a bunch of mechanics around which are eventually connected to some sensors whos values are usually fed into some kind of fly by wire system. which eventually figures out that a pitch up command is being input, so it then calculates how actuate the control surfaces to produce the desired effect. it seems like a ridiculous series of actions to basically have one computer talk to another. I guess I get that they are trying to produce an autopilot that can fly a plane without making modifications to said plane, but I would guess that purchasing new avionics and related components to add a traditional autopilot would be cheaper than an auto pilot packaged as an Android.

    1. I can see the usefulness of a robot that uses the human form factor – an android. But going to that much trouble only pays off when it can do many different things in place of humans. If it just flies an airplane and nothing else – even if it flies airplanes in general – it seems like a lot more trouble than an integrated autopilot system, which is far more efficient. Call me when it can change the baby’s diaper and cook dinner reasonably well.

    2. I agree that is is a very inefficient, error-prone and roundabout way of control input but due to the relatively slow reaction time of human pilots its a lot faster than required.
      The advantage is in the testing phase it can be safely used on any airplane. If it misbehaves the captain can just shut it off and take over control like he would with any first officer or student.

      Eventually when it works perfectly, the software will be certified and integrated in new airplane models bypassing the manual inputs.

  4. Not to go paranoid here but what if it was hacked ? Had an embedded command so that when it heard the right key phrase it would pull a 9/11 ? Or a future AI decides to go Colossus/Guardian and use the planes as weapons?

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.