Semi-automatic Aircraft Control

[shynung]

No, not what we usually call 'Autopilot' these days, but pretty close.

In the game Metal Gear Rising, a supporting character (a surgeon, presumably without any flight training) flies a heavy-lift helicopter to pick up a load of cargo that the protagonist plans to secure and obtain. The protagonists asks him whether he's fine flying the helicopter himself, which he replies that the common aircraft of that age has a semi-automatic mode, a technology spillover from unmanned aircrafts, that enables him to act as the protagonist's mission control right from the pilot's seat, while flying a heavy-lift helicopter.

So, the question is, how far are we from having that kind of aircraft control system? The one that can be piloted safely by an untrained, multitasking individual?

[Nuke]

i think we already have them. many commercial airliners can be landed entirely by computer. of course military drones have had this for years. fly by waypoint systems with point and click interfaces. now whether these systems work with an external load or in precision flight or whether its a good idea to use it for that or not, is entirely up for debate.

Edited December 26, 2015 by Nuke

[K^2]

We can build one of these, no problems. The pilot's skill is primarily in foreseeing how situation is going to develop, so he can correct for things in advance, as well as react properly in emergency situations. This is why you want a well-trained person in the seat, and why nobody bothered to enable a layman to fly with corrections from the flight computer.

[shynung]

Interesting. Though, I never see something like that advertised as a feature in aircrafts for sale to private owners.

Or is it actually pretty common? Common enough to be installed in most aircrafts that has computerized controls?

Also, is it possible to design a system safe enough to be flown by a layman? Sure, some hillbilly would eventually try to fly into something, but is it possible to make a system that would prevent that? Say, disabling all controls and fly the craft on computer until it reached a safe pre-defined attitude and altitude, then switch back to semi-auto?

[K^2]

I don't know of a lot of general aviation aircraft that are fly-by-wire. Naturally, the atutopilot can completely take over, but it can only do so much to "fight" the pilot. (Stick shakers, etc.)

Personally, I see two good options here. Either have pilot mostly responsible for control, and with flight computer, at most, filtering out "unsafe input" as some commercial aircraft already do. (I think Airbus does this on their fly-by-wire aircraft. Perhaps, someone will correct me.) In this case, the pilot is well-trained both in navigation and actual aircraft control, and there is no reason to do too much on the control correction. Or we go the self-driving car route and have the operator responsible merely for entering destinations. The flight computer would then be responsible for contacting ATC, filing flight plan, and following it safely to destination. There is very little reason not to go all the way here. There are no pedestrians in the air, nor terrain hazards. Most aircraft fly with transponders already. The few that don't are small, slow general aviation aircraft that tend to stay low and aren't too hard to avoid. Everything else is already being looked after by a number of computer systems responsible for air traffic. There is zero reason to let an unqualified person to even touch the stick.

[SomeGuy123]

The actual reasons have to do with liability, a limited market, the high cost of electronics properly hardened for aviation use, and a fear that such a control system might have hidden "edge cases" that cause fatal and expensive crashes.

For a quadcopter I've built one by purchasing a frame, the 4 motors, and an arduino and xbox controller RF module.  I also have an IMU, compass, a laser rangefinder, and a GPS module.

I had full auto flight, basically.  Think "videogame helicopter controls".  You could start the vehicle and just hold down the "up" arrow on one stick on the xbox controller.  It would keep itself level and adjust the internal "set point" for autopilot height.  The user did not directly or indirectly control the rotors.  

If you released the controls, the quadcopter would attempt to remain exactly where it was, subject to a bit of drift because GPS drifts.  

If you held down a direction key, the whole quadcopter would tilt and begin to fly in that direction, with the autopilot code what is actually still flying the aircraft, as it would only tilt over so far, keeping itself in stable flight.

Anyways, this is what "semi-automatic" helicopter controls would be like.  I'm well aware it would cost tens of millions of dollars to design the equivalent system intended for a manned, full scale turbine helicopter.  My little hobbyist system had a few edge cases I never fixed and all the electronics were just hand soldered exposed circuit boards on top of the copter.  You would have to use aviation grade connectors and shielded boxes and extremely expensive sensors and you'd need so much of them that it would weigh down the helicopter and reduce it's payload by 100kg or more.  (it's all the sensors and wiring and motor drivers and servos)

And you'd need a very rigorous formal development and testing regimen.  

But the long of the short of it is you could "hover" such a helicopter simply by releasing the controls.  Fly in a given direction by holding down a "direction yoke".  Don't worry about pushing the yoke down too far - the helicopter will never exit a stable flight region as it's autopilot is aware of where the boundaries are at all times.  Get in trouble, and you just release the controls, and the autopilot will restore the aircraft to a stable hover, assuming the actual rotors and engines and servos still work, recovering itself from all possible spins and other bad events.  (this is where the "edge case" fear comes in - there are high end quadcopter algorithms that can basically recover from anything, even deliberately removing a rotor, and the algorithms learn in realtime, but it's really hard to prove that there are absolutely no situations you have not thought of)

It would have things like belly cameras pointed straight down so you could land by just using the direction yokes so the helipad logo appears centered in the camera view, then hold down the "down" yoke to descend for a landing.  (the 8 controls would be Translation : Forward/Reverse/Left/Right/Up Down, rotation left and right.)

There would also be a "throttle" lever so you could just push the throttle forward and the helicopter would fly itself "forwards" at a relation % of the throttle (so 100% throttle means "fly forward as fast as possible given current conditions").  Like everything else, this throttle lever has nothing to do with what the engine is doing - you're just commanding the autopilot to fly forward, and it decides how much engine power to use.

Edited December 26, 2015 by SomeGuy123

[cantab]

As mentioned, the technology exists and is in every toy store quadcopter drone. The development and miniaturisation of this kind of flight assistance software and processors is why these small drones have only become widespread recently, whereas model aircraft and helicopters with conventional controls have been around for decades.

[Yourself]

1 hour ago, SomeGuy123 said:

Anyways, this is what "semi-automatic" helicopter controls would be like.  I'm well aware it would cost tens of millions of dollars to design the equivalent system intended for a manned, full scale turbine helicopter.  My little hobbyist system had a few edge cases I never fixed and all the electronics were just hand soldered exposed circuit boards on top of the copter.  You would have to use aviation grade connectors and shielded boxes and extremely expensive sensors and you'd need so much of them that it would weigh down the helicopter and reduce it's payload by 100kg or more.  (it's all the sensors and wiring and motor drivers and servos)

Most large helicopters already have all these things.  Most modern aircraft (even GA aircraft) in general already have a pretty sophisticated IMU anyway (the so-called Attitude Heading Reference System, i.e. AHRS).  The complexity and weight mostly comes from the servos used to drive the controls without pilot input.

Helicopter autopilots are definitely complicated, since there's a lot of cross-axis coupling in the different modes (changes in one control axis require changes in other axes to keep the helicopter stable).  The control dynamics also have to be tuned for different flight regimes as well (flying a helicopter in a hover is completely different from forward flight).  That said, helicopter autopilots aren't terribly uncommon.  They can hover, they can hold altitude, they can follow a GPS flight plan, they can do just about everything.

The major reason you probably don't see the sorts of video game controls is that it would be counter-intuitive to every helicopter pilot out there.  The skills involved flying that also wouldn't be the least bit transferable to a "dumb" helicopter (and "dumb" helicopters are what pilots usually first train on).  Also the fact that these sorts of systems are generally very expensive means there will always be dumb helicopters around because they're cheaper.

[shynung]

1 hour ago, Yourself said:

The major reason you probably don't see the sorts of video game controls is that it would be counter-intuitive to every helicopter pilot out there.  The skills involved flying that also wouldn't be the least bit transferable to a "dumb" helicopter (and "dumb" helicopters are what pilots usually first train on)

Why is this the case? Does this apply to fixed-wing aircrafts?

[*Aqua*]

4 hours ago, K^2 said:

Either have pilot mostly responsible for control, and with flight computer, at most, filtering out "unsafe input" as some commercial aircraft already do. (I think Airbus does this on their fly-by-wire aircraft. Perhaps, someone will correct me.)

Afaik there are quite a lot of people who criticize Airbus for their flight computers. The argument is that they limit the pilot too much.

Some of you probably remember the near crash of an Airbus: https://www.youtube.com/watch?v=z42fchrzhHY
When an Airbus lands the flight computer switches to ground mode and in this mode maneuvers like touch-and-go won't work. The flight computer will prevent that to avoid accidents. The plane already touched the ground and the computer switched the mode from flight to ground when the gust came. And that's why the pilots couldn't prevent the wingstrike. They needed to shut the computer assistance completely off to regain control and take off.

My point is are flight algorithms good enough to react on all situations in the right way? Are the systems secure enough to prevent unauthorized manipulation from the outside? Are they reliable enough that they can safely land a plane even if parts of the computers are destroyed (fire, power loss, ...)? I know that commercial planes usually have three flight computers which control each other, but what if two of them don't work right? Then you have three computers with three different interpretations of the current situation.

 

45 minutes ago, shynung said:

Why is this the case? Does this apply to fixed-wing aircrafts?

Think about it: What happens if you push the stick of a "dumb" helicopter? It'll tilt forward, the flight vector goes down, you'll start to lose height, you may need to change the AoA of the blades, etc. A "semi-auto" helicopter will do all that for you which means you'll never experience that and never learn how to react on that.

When learning how to drive a car in my country you have two choices: Learning driving in a training car with manual transmission or one with automatic transmission. If you choose automatic (it's cheaper) you are not allowed to drive cars with manual transmission. Why? You might need to shift gears in a critical situation and if you fail at that because you never really learned it there's a higher chance that you'll fail. A fail with a weapons that weights one or two tons isn't exactly healthy.

It's the same with helicopters and planes. It's okay if there are some assistance computers to lift the load on pilots but nothing beats a pilot in situations software can't handle. 

Edited December 26, 2015 by *Aqua*

[Shpaget]

http://ardupilot.com

A free, open source autopilot that can drive cars, fly planes, helicopters and multirotors.

It can take off throw assisted and land autonomously, fly a route via 3D checkpoints and quite a bit more. It is designed to fly models, but there is no reason it couldn't fly a plane of any size.

Commercial airliners have autopilots that fly along the route, steer and change throttle as needed. Waypoints are added and removed as needed during the flight, but the majority of route is loaded before the plane takes off. When the pilot dials in the change in altitude (as cleared by ATC), which is given as target altitude and rate of climb/descend, the computer does its best to make the plane fly accordingly.

During cruise, pilots rarely fly the plane. Autopilot is turned on almost all the time, since it is much better at holding altitude than humans.

Edited December 26, 2015 by Shpaget

[shynung]

10 minutes ago, *Aqua* said:

Think about it: What happens if you push the stick of a "dumb" helicopter? It'll tilt forward, the flight vector goes down, you'll start to lose height, you may need to change the AoA of the blades, etc. A "semi-auto" helicopter will do all that for you which means you'll never experience that and never learn how to react on that.

Actually, I know that's the cyclic stick, and I know what it does. But I get your point; a semi-auto helicopter would probably read that as a 'translate forward' command, and compensate any altitude changes through blade AoA (collective).

[SomeGuy123]

1 hour ago, *Aqua* said:

My point is are flight algorithms good enough to react on all situations in the right way? Are the systems secure enough to prevent unauthorized manipulation from the outside? Are they reliable enough that they can safely land a plane even if parts of the computers are destroyed (fire, power loss, ...)? I know that commercial planes usually have three flight computers which control each other, but what if two of them don't work right? Then you have three computers with three different interpretations of the current situation.

The other side of the coin is that human pilots make certain common mistakes.  You can prevent these common mistakes with a computer in the way, at the expense of making it more difficult to recover from rare mistakes.

Analogous to how an automatic transmission sometimes picks the wrong gear, but prevents you from grinding the gears or stalling your engine.  

Anyways, by blocking common mistakes, you may prevent more accidents than you cause by making it harder to save an aircraft when a rare edge case happens.  

There are so many confounding factors, and lost aircraft are so rare, that I don't know if in the end there is any statistically significant safety difference between Airbus and Boeing as a result of this philosophical difference, but considering doing it this way is not a bad idea.  It's a valid engineering tradeoff.  

In the long run, as you keep making the software better and covering more of those rare edge cases, you would expect that you would come out way ahead, because human pilots are a lot harder to make universally reliable than software.

Edited December 26, 2015 by SomeGuy123

[wizzlebippi]

NASA has looked into this in their highway in the sky program. The goal at the time was to develop an autopilot capable of flying the aircraft from takeoff to landing and see if the average joe could make it all work. Supposedly their work was the basis for the part 23 technically advanced aircraft category and most modern GA glass cockpits.

Also, aircraft is plural. Had a professor that threatened to fail anyone who said or wrote aircrafts.

[Pixel of Life]

On 12/26/2015 at 3:29 PM, K^2 said:

Personally, I see two good options here. Either have pilot mostly responsible for control, and with flight computer, at most, filtering out "unsafe input" as some commercial aircraft already do. (I think Airbus does this on their fly-by-wire aircraft. Perhaps, someone will correct me.)

Yes. It's called "Flight Envelope Protection". The Airbus FBW won't allow any kind of dangerous maneuvers - accidental or intentional - unless the pilots either completely switch off the Flight Control System or somehow put it in one of the Alternate Law modes where some of the protection systems are offline. Newer Boeing aircraft also have such a system but it's only designed to prevent pilots from accidentally doing dumb stuff. It can be overridden by applying enough force on the controls.

[mrfox]

On 12/26/2015 at 3:34 PM, shynung said:

...common aircraft of that age has a semi-automatic mode, a technology spillover from unmanned aircrafts...

So, the question is, how far are we from having that kind of aircraft control system? The one that can be piloted safely by an untrained, multitasking individual?

Right at the beginning actually

http://www.aopa.org/News-and-Video/All-News/2012/December/19/Diamond-to-offer-auto-landing-in-2016

http://www.aopa.org/News-and-Video/All-News/2015/September/22/Diamond-Debuts-Autoland-System

This development stems from Diamond's work in unmanned versions of their small single and twin engined aircraft, resulting in some of the first fly-by-wire light aircraft that will be coming on the market in the next few years.

Edited December 29, 2015 by mrfox

[GoSlash27]

Actually, autonomous helicopters have been a thing for a while now.

Best,

-Slashy

[YNM]

IRL large helicopters have more advanced balance systems compared to those small ones. Heard this over from SED where he took helicopter lessons and the instructor/trainer says that larger helicopters are easier to fly than simpler smaller ones thanks to semi-automatic balancing between the collective, cyclic and the pedal. Also, flying helicopters in straight line is far easier than hovering over one spot and not going anywhere else, he noted.

CORRECTION : Actually, not dampening. The reason why larger helicopter are "easier" to stabilize is because they're less sensitive (thanks to hydraulic non-direct input says wiki) and have augmentation, purest (and therefore hardest) helicopter are small ones like R22 or so. More likely to overcorrect in small heli than large ones.

Edited January 1, 2016 by YNM
clarifying a bit late

[shynung]

13 hours ago, GoSlash27 said:

Actually, autonomous helicopters have been a thing for a while now.

Best,

-Slashy

That's really cool. Something like this would really help in shipping and construction as aerial cranes.

I'm guessing it's just a matter of time until we get fully-autonomous personnel transport craft, the kind we operate by merely entering addresses or GPS coordinates.