17

If a plane can already autoland and stay aligned while roll-out by itself, on a CAT-3 ILS-Approach, I asked why there is no auto-takeoff-function on todays autopilots.

Because it should be technically possible and much safer than a hand flown takeoff.

  • For the technic: When on the runway, the pilot presses an auto-take-off button. The plane stays aligned by following a normal CAT-3 Localizer. Takeoff-Thrust is made with autothrottle. At Vr, the plane rotates by itself, smoothly,and activates VNAV and LNAV and follows the SID. It is also possible to implement a live wind sensor to dynamically update V-Speeds.

  • For safety: In case of an emergency during take-off, a computer take-off would be more safer. The plane knows the speed and runway length. So in case of an emergency, it knows if it's below V1 and also if the available runway length is sufficient for a safe rejected take-off. A computer always has a lower reaction time than humans and they can response faster. A computer can immediately decide wether to take off or reject based on lots of live data. A (rejected) take-off performed by a computer should be times safer.

So, why has an autotake-off-function not been developed yet? And what arguments would speak against an auto-take-off function?

-- There is not an answer to my question here: On modern commercial airliners, how much of the flight could be fully taken care of by the auto pilot? as stated below. I want to know what reasons, if there are any, speak against auto-take-off.

Community
  • 1
Noah Krasser
  • 8,992
  • 3
  • 50
  • 100
  • 3
    What happens when a baggage train rolls across the runway? Or a deer? – Ron Beyer Jan 17 '17 at 14:50
  • 2
    @RonBeyer the same as what happens on landing surely? – Notts90 Jan 17 '17 at 14:55
  • 2
    @RonBeyer Good Point, but I think pilots should still pay attention to whats happening in front of the plane and should still be able to reject the take-off. Aircraft could also build in a radar which detects obstacles on the runway and reject the take-off. – Noah Krasser Jan 17 '17 at 14:59
  • 2
    For all the same reasons pilots don't always use autoland – FreeMan Jan 17 '17 at 15:38
  • 4
    Possible duplicate of On modern commercial airliners, how much of the flight could be fully taken care of by the auto pilot? – FreeMan Jan 17 '17 at 15:40
  • 6
    Where comes this belief that computer are always right? – Peter Jan 17 '17 at 16:22
  • 1
    @Peter If all sensors are working properly and the computer is programmed properly and working, then it should be always right. – Noah Krasser Jan 17 '17 at 16:24
  • 3
    @Peter they may not always be "right" in the human sense of the word, but they definitely dont make human errors. – Jamiec Jan 17 '17 at 18:06
  • 1
    When an emergency happens, I'd rather have pilot that's taking over from the computer who's done the procedure a couple hundred times in the last year than one that has watched a computer do it and practiced in a simulator. Computers make human errors all the time. At least until they start programming themselves. I'd rather the computer took over if the pilot fails due to illness or death than the other way round. Call me a Luddite if you like. – Greg Taylor Jan 17 '17 at 21:21
  • There have already been autonomous flights, particularly with military drones, so I would have to answer yes, the technology exists.... – Lnafziger Jan 18 '17 at 01:31
  • 1
    @GregTaylor: That's is the point and as a programmer I know that programms are full of errors and situations they can't handle (both can be improved with time and work). And if they program themselves we are either in trouble or in a better world, I can't say which one. – Peter Jan 18 '17 at 10:50
  • 2
    @RonBeyer The same as happens when a baggage cart or deer runs out in front of a self-driving car. Computers aren't perfect but they're usually better than humans at this kind of thing. – David Richerby Jan 23 '17 at 08:35
  • @NoahKrasser the computer always gives the correct answer according to its programming, but its programming will NEVER be perfect. There's always things you don't anticipate when creating software, however careful you are, and there are always things missed in testing, however meticulous you are. I've been working in software engineering for over 2 decades, going on 3 if you count my time as a student, and that's the one common. – jwenting Nov 22 '18 at 06:19

2 Answers2

19

With some refinement and programming, the technology is easily there, sure; it just needs a few million dollars of design and certification costs for very little benefit.

The main reason we have autolands is because it saves airlines money, by allowing the plane to land safely in minimal visibility instead of diverting to another airport. Aircraft taking off don't have this problem as they can do it in just about zero visibility. If there is no commercial incentive, the reality is it is not going to happen. Only when pilots are considered surplus to requirements will this ability be developed in earnest.

There could be an argument made that it should be done for safety's sake, but I don't buy it. Pilot errors made on the takeoff run, resulting in an accident are incredibly rare. Maybe an auto takeoff could prevent a few tailstrikes but I can't see much more benefit.

Whenever you introduce something new you have to think long and hard about the ramifications of it. There's a whole discussion to be had on how rejected takeoffs would work - generally speaking, you do not want automation doing something that the pilots aren't expecting. It is sometimes better to take an issue into the air than take the risk of a high-speed RTO even under V1, how will the computer judge that when there are few hard and fast rules for it? Do you trust the human programmers to account for every single scenario? Then there's pilot competency, how do they keep their skills up when they could go months without flying the aircraft? If they are in the habit of delegating their decision making to the computer, will they be able to handle the heat in that career-defining moment when the computer fails and it's all on them?

Ben
  • 14,065
  • 3
  • 47
  • 73
6

It is perfectly possible- in a very limited sense and has been demonstrated in aircraft like Global Hawk (not a commercial airliner, but it can be done here). The problem is what to do when the unexpected happens.

Technically, it is possible- you can set the TOGA thrust and set rotation speed and let the aircraft do the take-off, but the problem is, that's about it. What will the aircraft do when someone else decides to crash the runway? It has to detect, decide and act appropriately- which increases the system complexity multiple times due to the sensors required (you are already asking for a radar to detect obstacles; the sensor count will go up, not down).

Safety wise, the issue is, the aircraft can react to known problems. What will it do when something that's not in its library occurs? Unlike human pilots, they cannot experiment and compare notes with the other crew and come to a decision. That computers are better and safer than humans is based on the belief that it has access to all data and is programmed to do the correct thing- which is not true always.

Another important thing during takeoff is that the flight crew is in constant communication with the ATC; by far this is the most complex part in the actual take-off barring something unexpected. Computers are not capable of doing it right now. Also, they have to respond to emergencies- both internal (system failure etc) and external (changing ATC instructions, traffic etc).

Also, what will you do when the computer decides to reboot itself in the middle of the takeoff procedure?

aeroalias
  • 100,255
  • 5
  • 278
  • 429
  • 3
    What will it do when something that's not in its library occurs? Of course the pilots should still be able to perform a rejected takeoff on their own decision. And if ATC asks to reject the takeoff the pilot can also do it on their own. Also, what will you do when the computer decides to reboot itself in the middle of the takeoff procedure? You can ask this question for all avionics in a modern aircraft like Autopilot, EFIS, etc. – Noah Krasser Jan 17 '17 at 16:40
  • 7
    All of your arguments could equally be made against auto land. I don't see the difference. – Jamiec Jan 17 '17 at 18:08
  • 2
    Unlike human pilots, they cannot experiment and compare notes with the other crew and come to a decision. I'd like to know what human pilot has the time to compare notes with other pilots in the middle of a takeoff. – Vikki Feb 20 '19 at 22:08