1

What is going to change with auto-pilots on commercial aircraft in the next decade or so? Will airlines allow an auto-pilot to control an aircraft when it lands and takes off? Will we start using

Kenzoid
  • 119
  • 3

2 Answers2

1

A decade is not a long time at all.

What the future holds for the next decade is the new aircraft-ATC communication system, CPDLC. All of the focus is on that. Enabling data links between pilots and ATC is a hard prerequisite for any significant autopilot development.

It's very possible and likely that nothing revolutionary will happen in autopilots.
It's possible and likely that something more minor is going to change.

Some believe autopilots will get more sensor and control fusion. Still nowhere near full control. Others believe more manual flying will become standard procedure, to reduce the damage when the electronics fail and reversion to direct law is required.

No one knows that for sure, so it would be pure speculation. What is known is that autopilot development will not be the next decade's primary focus. For any major new system to fly passengers in 10 years, it would have to be in advanced testing stages already.

Therac
  • 26,835
  • 2
  • 73
  • 110
1

99% of the time pilots spend on an airplane is basically automated.

Flying a strait line doesn't take much input other than making minor adjustments to account for wind direction ect...

Landing and takeoff however are the most important and difficult aspects of flying. Precision input is constantly applied to the controls with much less forgiveness when a mistake is made as there is no time to correct for it.

However computers are already very good at doing these procedures. In normal conditions it wouldn't even be a problem for a computer to land these planes. But that's the key, normal conditions.

What about an emergency? What about high wind gust? What about a hijacking? A birdstrike? When something goes wrong and programmers had no way of foreseeing the problem no computer could generate algorithms on the fly to deal with it. A trained pilot however just might be able to save the day.

enter image description here

The airline industry is very very strict about risk management and any changes to be considered must undergo extremely hard testing.

If a company came out tomorrow with an actual working piece of software that could fly a plane to its best possible abilities in any condition, and out perform humans in every single possible situation I'd reckon it would still take 10+ years of red tape just to start human trials on it.

And fyi we do not have that type of tech today. There have been great strides recently in terms of autopilots using techniques such as machine learning, but these practices basically started in the past decade and the knowledge we have is very young.

Self driving cars require human input anytime it gets in a situation it doesn't understand. A friend of mine owns a tesla and said his car requests input every 15 miles on average while driving on the outskirts of the city.

I tried to find better stats for an average but didn't see any from a quick google search, still though even his case of required input every 15 miles shows we have a way to go.

How much further do we have to go? I don't know. I don't think anyone does. It's like asking what kind of smart phone will we have in 20 years. Will we even have smart phones?

But one things for sure, before it would ever be allowed to happen, if ever, it would have to have a fail/pass ratio of basically 100%.

YAHsaves
  • 1,211
  • 1
  • 9
  • 21
  • 99% of the time on autopilot would leave just 90 seconds for takeoff and landing combined on the average 2.5-hour flight. That's if we don't count taxiing. – Therac Jul 25 '19 at 18:20
  • @Therac Ha ha you got me there, but it's a valid point. Pilots taxi planes on the ground, so any "replacement" autopilot would have to also be a self driving ground vehicle also to navigate through modern airports. – YAHsaves Jul 25 '19 at 18:24
  • 1
    @Therac The exact limits depend on company policy, but it's not uncommon for the AP to be allowed from 500ft AGL after takeoff until 500ft AGL before landing. At the speeds airliners fly, that's probably less than one minute on each end, including ground roll. 99% is actually possible. Still, the tiny fraction where the humans are flying is the by far most challenging and where things tend to go wrong; I've read 97% of accidents happen within 2 miles of an airport. – StephenS Jul 25 '19 at 18:33
  • @StephenS Who in their right mind would fly autopilot for climb and descent if they're allowed to fly manual? A lot of people, really, but most join the profession to fly... It's company policy and fuel economy that force more autopilot time. If fuel gets cheaper than reputation (lost when pilots with no manual skills end up making mistakes), even a mild reversion of that trend is possible. – Therac Jul 25 '19 at 18:54
  • @Therac That would make a good new question; the answer is far too complicated for comments. I was only addressing the 1% issue. – StephenS Jul 25 '19 at 19:01