Recently someone on another thread recommended the following video:

Andrew Godwin - What can programmers learn from pilots? - PyCon 2015

where the speaker, at about ~3:14, talking about pilots mentions:

not to the point where we can replace them with computers, because we're not there yet

which got me thinking:

  • How close are we to replacing pilots with computers? (like totally)
  • $\begingroup$ Technologically - a very long way off. Politically -even further! I doubt it'll happen in any meaningful way in my lifetime, personally. $\endgroup$
    – Dan
    Aug 25 '15 at 14:24
  • $\begingroup$ For cargo flight, I woul'dnt be so sure. $\endgroup$
    – Antzi
    Aug 25 '15 at 14:29
  • $\begingroup$ Several businesses are working on autonomous aircraft including passenger-carrying aircraft. One is Astraea $\endgroup$ Aug 25 '15 at 15:06
  • $\begingroup$ I wouldn't trust any computer to fly my airplane. Most likely their will be unmanned drones in the future, but I don't think a computer controlled aircraft will ever happen $\endgroup$
    – Ethan
    Aug 26 '15 at 21:29
  • $\begingroup$ In principle, an autonomous combat aircraft capable of carrying out air to air and ground attack missions is technically feasible, and could in fact be produced using mainly existing components from other systems such as cruise missiles (for the navigation system) and smart munitions (for the target identification systems and armament). $\endgroup$ Jan 25 '18 at 12:28

As a person that "tries to replace pilots with computers" I personally say: not even close.

  • The processing power (certified for flight, not the latest pentium sitting under your desk) is too behind, in particular concerning the point here below.

  • Synthetic vision is still extremely lacking. This is currently the biggest hurdle to overcome. While pilots can look out of the windows and assess the situation (with some limits, admittedly), computers still have several problems with it, starting from a general lack of cameras and going all the way to the daunting level of processing power required.

  • Ability to adapt is still a long way to come. Computers are "fundamentally stupid", they do what they are told (programmed) to do, they are not capable of adapting to an unforeseen (and never experienced before) emergency. There are several people working in machine learning and the field has advanced significantly in the latest years, but good luck certifying those programs.

  • 1
    $\begingroup$ Google has yet to buy Boeing or Airbus. This is a matter of time. $\endgroup$
    – mins
    Aug 25 '15 at 22:04
  • $\begingroup$ @mins google is not exempt from certification $\endgroup$
    – Federico
    Aug 26 '15 at 4:50
  • $\begingroup$ Nor is Google exempt from physics $\endgroup$
    – FreeMan
    Aug 26 '15 at 15:02
  • $\begingroup$ to whomever has downvoted: care to explain? $\endgroup$
    – Federico
    Sep 17 '15 at 13:21
  • $\begingroup$ A lot of people seem to assume that a tactical UCAV would necessarily have to replicate every capability of the manned multi-role fighter to be viable, when the real question is one of how much capability it needs to have in order to be cost-effective. $\endgroup$ Jan 26 '18 at 20:57

This is a complex question and one that involves both a technical aspect as well as a legal aspect.

From one standpoint there is nothing stopping the FAA from requiring that a human pilot remain in the plane should anything go wrong (or simply to create jobs). Lobbying power should not be underestimated here, there are lots of pilots (and a union behind them) that would be out of work.

But lets talk machine take over for a minute. In some regards parts of the systems do exist. Obviously autopilot to an extent can fly a plane but this generally involves some human interaction and generally holds speed/altitude/heading a change in any one of those may require physical control changes. In a similar vain auto-land systems do exist and they do work. HOWEVER they are complex to set up and require human interaction to currently be used. As far as I know there is no automated system that can currently control takeoff (i could be wrong about that though). Do we possess the pieces to assemble a self flying plane, possibly... To do so, and to get it FAA certified is a very big topic.

The other issue that arises is updating the whole ATC system. If you have not been following the news this is already a mess. But lets say you had a plane that had no one flying it. If ATC needed to reroute the plane, or issue a change in course it would need a system to do so. Currently that is handled via a radio and person to person communication. You would need to dispatch some kind of command to the plane via a data link to do it in an automated system. Then again you can automate ATC etc. etc. (lets not go down that road.

  • $\begingroup$ A lot of people seem to assume that a tactical UCAV would have to replicate every function of the manned multi-role fighter to be viable. But a UCAV which only had the capability of flying to a pre-determined location, dropping some mines and then returning to a base location might be militarily cost effective. Adding an infrared sensor and smart munitions to enable it to attack armoured vehicles directly would be quite simple. $\endgroup$ Jan 26 '18 at 21:17
  • $\begingroup$ @J.Southworth why do you think we are talking about "tactical UACV"s? $\endgroup$
    – Federico
    Jan 28 '18 at 12:33
  • $\begingroup$ There was nothing in the original question to indicate that it didn't refer to military aircraft in tactical roles, so I assumed that it did. $\endgroup$ Feb 16 '18 at 18:08

How close are we to replacing pilots with computers?

In some ways, this has already happened in some limited areas.

Flight Engineers and Navigators

As a precursor to replacing pilots, we have already replaced non-pilot crew on the flight deck. The flight engineer role has been replaced largely by computerized engine systems.

A significant problem for the remaining pilots can be maintaining their skills during their time in the cockpit when they spend long hours mostly monitoring systems rather than doing hands-on piloting.

I suspect, a lot of what commercial and military pilots do boils down to issuing commands to computers, rather than manually (if indirectly) moving control surfaces.

Bomber pilots

In the first gulf war, 297 missions were flown by autonomous aircraft. Many explosive payloads were delivered to targets by small aircraft that flew hundreds of km. These navigated and flew without any on-board or remote human pilot controlling them. Cruise missiles.

The alternative would presumably have been greater numbers of human pilots in conventional bombers dropping smart bombs over the target.

Ground attack pilots

Some jobs that would have been carried out by a pilot inside a ground-attack aircraft are now carried out by a UCAV with a remote pilot. These UCAVs have increasing amounts of autonomy. Some are intended to have full autonomy. It isn't hard to envisage a not too distant scenario where the main human interaction with the aircraft is not piloting it, but authorizing firing of weapons.

Helicopter pilots

Jobs like power-line inspection are, in some remote areas, mostly carried out from helicopters with human pilots but we seem to be on the brink of replacing some of those pilots with drones. That seems a few technically short steps from using autonomous drones. The main time-consuming barrier is probably regulatory change.

Currently, to operate a drone for this purpose, you don't need to be a licensed pilot. Is such a drone operator a pilot. Have they replaced one?

(like totally)

As others have answered, nowhere near.

  • $\begingroup$ In technological terms the cruise missile is the closest relative to the autonomous UCAV. $\endgroup$ Dec 24 '17 at 13:51

Not the answer you're looking for? Browse other questions tagged or ask your own question.