I read about how the MiG-29 fighter jet has a button in the flight controls that, when pressed, returns the aircraft to straight and level flight no matter what attitude or angle of attack it was at previously. Seems like this would be a great control button for disoriented or panicked pilots (I.e., Air France 447) to have; is it practical for airliners?
I can think of so many things that it would have to be ready for and in a no power situation, straight & level is not going to work for many seconds.
Can it be overridden?
$$$$$$$$$$$$$$
IMO, it is a solution looking for a problem that is so rare in airliners that I vote no.
Fighter pilots need it as they G (intertial force trying to stop them turning corners) themselves silly… no blood in the brain …
Passenger pilots can just press “autopilot on”.
Air France didn’t do that as they believe the instruments were reading faulty readings. ( If they looked at GPS they wouldn’t have to worry the instruments were affected by ice, and other blockages or just simple mechanical failure.)
I don’t think the autopilot ON button will have any effect in situations where you actually need it. An unusual attitude generally needs much more assertive control inputs than the autopilot will do. That said, the Airbus fly-by-wire philosophy basically has all of the pilot inputs being interpreted by a pseudo autopilot anyway… in normal law, so maybe an Airbus autopilot would recover from a UA. I don’t know, outside my area of experience. Certainly the BAe146 autopilot will throw its metaphorical hands in the air long before a half competent pilot would. I recently hand flew an approach and landing because the turbulence was too much for the autopilot, it wouldn’t stay connected.
I don’t think a straight and level button would have helped Air France as some of the airspeed data had become unreliable.
On the subject of using the GPS to check speed, this is of limited use. What’s important is the indicated airspeed, and this is only given directly by the airspeed indicator. I might be cruising with 260 knots indicated on the ASI while our true airspeed is 400 knots and GPS might show a ground speed of anywhere between 300 and 500 knots. GPS derived ground speed also won’t account for a descent angle. This doesn’t matter for normal operations but if you find yourself in a vertical dive doing 400 knots, the GPS will show a ground speed of 0. If you have the presence of mind to make allowances for the GPS limitations then you probably wouldn’t be in trouble in the first place. A basic power setting and pitch attitude is all you need to counter unreliable airspeed data. This is not a “memory item” on the 146 but I know if I set around 88% N1 on the engines and an attitude of 5º nose up I will be safe while my offsider gets the checklist for unreliable airspeed.
I don’t have a problem with the idea of a “fix my mistakes” button, but you’ve entered the realm of the autopilot acting as a backup for the pilot when the current philosophy is the opposite.
These systems, called Pilot Activated Recovery System (PARS) or Auto Ground Collision Avoidance System (AGCAS) is already installed in some F-16s and it is also planned for the F-22:
http://www.nasa.gov/centers/armstrong/Features/Auto-GCAS_Installed_in_USAF_F-16s.html
Rockwell has proposed this for some commercial jets, sometimes called a “digital parachute”: http://www.wsj.com/articles/SB10001424052702304791204576402083915084432
Whether it would help cases like Air France 447 is less clear. That was initially caused by a degraded flight control system, so the system itself has already been deprived of reliable airspeed data. However the inertial platform (ie “artificial horizon”) was working, as was the altimeter. There was sufficient data for either a human or an automated system to avoid stalling.
If an auto-recovery system existed on AF447 but they invoked it after they were in a deep stall descending at 10,000 feet per minute, it would be trickier. Maybe it would work. However this would require the pilots to trust the system and not fight it.
In the case of AF447 there was total breakdown of cockpit discipline and both pilots were fighting for control of the aircraft without telling the other. Since Airbus fly-by-wire controls are not mechanically linked one pilot cannot physically feel if the other pilot makes control inputs. Each joystick has a “takeover” button to lock out the other pilot, which only lasts until the opposing pilot hits his button to take back control. Flight recorder data indicates they were repeatedly fighting each other for control of the aircraft without understanding what the other was doing.
In such a situation it is possible the pilots would fight the automated recovery system, whether it was manually or automatically invoked.
In the specific case of AF447 it would have failed in the early stages of the mishap because the instrument data sources were malfunctioning. The same ones the magic “save the jet” software would have been referring to determine aircraft state.
In the later stages after the pitot blockage had cleared it might still have been fooled by the bad design of the stall warning system. Avoiding stall would have been one of the prime directives of the system and once they were fairly deep stalled the entered a paradoxical software glitch where the worse the were stalled the happier the stall warning system was. And the closer they got to recovery, the more the stall system warning protested they were screwing up.
What the pilot(s) lacked was a clear understanding of their predicament. At the present state of the art a software “save the jet” system as built by Airbus would probably have made a worse mistake.
Could such a thing be invented and work for many scenarios other than AF447? Sure. How often do those scenarios come up? Very, very rarely.
Might be more useful to a Drone.
That is not exactly correct. When reliable airspeed indications were lost, the correct pilot action was maintain proper pitch profile and throttle. That could easily have been done by either a human or a computer, and in fact human commercial pilots do that many times.
In the AF447 case the autopilot was not designed to continue when airspeed indications were lost, but it could have been. Alternatively it could disconnect and give the pilot the option of pressing an “emergency maintain level flight” button. These are just software and procedural design decisions. The data was available to permit this.
The avionics knew the plane was stalled. Due to a poor human interface design, it confusingly suppressed then presented the warning – but the system had access to the correct data. A “return to level flight” mode would not have been confused by the inconsistent warning presented to the pilots because it would be looking at the original data.
That part is true, and even if implemented it would require rational behavior by the pilots. There have been numerous cases where pilots disregarded or overrode warning systems like stick shakers and stick pushers and stalled the plane anyway. These infrequent conditions happen when there is total confusion in the cockpit and a complete breakdown of coordination. Such a mental state could also prevent a “return to level flight” system from working.
If the pilot failed to press the “return to level flight button” or did so then overrode it, the plane would still crash.
With both AF447 and AirAsia 8501, the inertial platform and primary flight display were working perfectly. In both cases the copilot pulled back on the sidestick for long periods, despite the big PFD showing blue sky. The screen is painted blue for sky and brown for ground to make it immediately obvious when the plane is pitched up or down. If a pilot is going to sit there looking at the blue sky on the PFD for several minutes while he’s pulling hard back on the sidestick, it doesn’t take a genius to guess what will happen.
However current automation design is highly conservative because of the risks you stated. There is always the possibility a sensor failure could cause automation to make incorrect decisions, thus the system is “spring loaded” to return control to the pilot.
But in both AF447 and AirAsia 8501 cases, there was sufficient reliable instrument data for an automated system to maintain control, if designed for that.
I’m confused. I thought this is exactly what an autopilot did. If the pilot takes his hands off the controls, the plane will automatically return to level flight. Is this not correct?
No, the pilot typically does not handle the controls during autopilot flight. If the autopilot is engaged and he applies a certain amount of control force, it will automatically disengage without him having to manually hit the disengage switch. The autopilot will then not re-engage unless he manually connects it.
The autopilot is not specifically designed to maintain level flight. It is designed to fly the plane as programmed. E.g, it is common for commercial pilots to engage the autopilot and autothrottle shortly after takeoff, and the entire flight including landing is done by autopilot. It is obviously climbing, cruising, turning, descending and landing under autopilot control, not just maintaining level flight.
There’s a lot of confusion about what autopilots do, and how they work. Many people seem to think pilots just flip the switch and then sit and do nothing. It actually takes a fair bit of know-how to make aircraft automation do what you want when you get into modern avionics.
As a simple example, let’s say you want to climb in the light jet I fly. You have your choice of two modes: Do you want to climb while holding a certain airspeed, or would you rather select a certain climb rate? Either way, you’ll have to also set power since my plane doesn’t have auto-throttles. You would also have previously selected a lateral mode (meaning, where the plane is going on the map) before engaging vertical.
A more complicated example: You want the plane to fly a charted arrival procedure AND follow the associated altitude step-downs. In my plane you would do the following:
- Load the arrival through the Flight Management System (FMS)
- Verify the altitude step-downs against the chart, then select the lowest point in the Altitude Pre-Select Window
- Select “Nav” in the AP to follow the waypoints
- Select “VNav” (vertical navigation) in the AP to fly the altitudes (and adjust throttles each time)
That’s if you’ve been cleared “via the approach” from one of the first fixes on the procedure. Sometimes you don’t get cleared down all the way, or have to enter the arrival from an intermediate fix, which requires more FMS inputs. Later on you’d program an approach, which engages still more AP modes depending on what you’re doing.
So we’re kept rather busy programming and monitoring the automation. AP is a tool, and it takes work to utilize it efficiently.
“We?” Llama, where’ve you been when all the other pilots were posting their heads off? Good to know ya.
I always read the aviation threads, but LSLGuy and a few others usually beat me to whatever point I was going to make. And very likely better than I would have made it. 
I forgot to add, even if the autopilot is set for level flight it does not have the capability to recover the aircraft from “unusual attitudes”. E.g, if the aircraft is stalled, spinning, etc.
While it is not technically impossible for automation to perform this, as stated previously it has to be a specialized design such as the system being put in the F-16 and F-22. You would normally want that separate from the regular autopilot, because it would be making decisions with possible degraded sensor data. It would be a step of last resort, hence the term “digital parachute”.
In general the automation design philosophy is handover control to the pilot in case of sensor malfunction or unusual flight attitudes. The thinking is in those extreme situations a human is best equipped to handle it.
Also some commercial airliners are not flight tested outside the normal envelope, so there is limited data on which to base a computer model for either simulation or recovery from unusual attitudes. By contrast fighter planes are tested outside the envelope. Here is a photo of a spin recovery parachute on an F-16 and F-22 which the test pilot can activate to break the aircraft out of an uncontrollable spin in flight testing. They don’t flight test airliners like this, so the data is not available to model aircraft behavior in that regime:
F-16 spin recovery chute: http://www.dfrc.nasa.gov/Gallery/Photo/F-16AFTI/Small/ECN-22193.jpg
F-22 spin recovery chute: https://sites.google.com/site/quicklinkall/f22-raptor-06.jpg
This system behavior was a factor in the 1994 crash of Aeroflot flight 593.
But pilots might use it multiple times in non-emergency flights too. Anytime you want to level off and cruise.
In many AP systems this is quickly accomplished by selecting “heading” and “alt”. At least in my plane, that would de-select any other modes and arrest a climb or descent. Not often that I need to do this, just once in a while.
An autopilot can already do that – for normal flight. It is called “altitude hold”. If the autopilot is engaged with ALT hold on, it will hold that altitude. It can also climb or descend at a specified rate and then hold a predetermined altitude or will hold the current altitude anytime the pilot engages ALT hold. It’s one button.
In general this is not needed since commercial flights typically have the autopilot engaged the entire flight except for takeoff. A sophisticated Flight Management System follows the course based on a series of waypoints:
Current autopilots are not designed to recover from spins, stalls, or other unusual attitudes, or with degraded sensors.
If an emergency “return to level flight” function was added, it would likely not be used in non-emergency situations since the autopilot can already do that.
It would use special algorithms which take into consideration unusual attitudes and degraded sensors. Using it would entail risk since those things are difficult and dangerous to test. In some airliners they are never tested – even during instrumented developmental flight testing. So there is no good data to base such a recovery algorithm on.
Even in an FAA Level D flight simulator with a motion platform having six degrees of freedom and high-def realistic panoramic visual systems, they cannot always fully simulate “out of envelope” aerodynamic behavior since that data was never obtained during flight testing: http://www.cae.com/uploadedImages/Content/BusinessUnit/Corporate/News/2013/images/CV7478Sim1.jpg
If AF447 had had such a button, do you think the pilots would have used it? They didn’t think they were disoriented; the only things they were confused about were their angle of attack and airspeed, which were so extreme that even the plane’s computers were confused about what was happening.
In order for a PARS button to be useful, two preconditions need to be met:
- the pilot must actually be disoriented, and
- the pilot must recognize that he is disoriented and in a situation that requires immediate correction.
Fighter jets are capable of violent maneuvers that may indeed leave the pilot truly disoriented, in which case those two preconditions will be met. Rather than study the instruments for a few seconds or try to find the horizon outside the window, PARS may be a lifesaver, especially at low altitude.
OTOH, how many commercial airliner crashes have met these preconditions? I suspect very few, given that these aircraft tend not to maneuver as violently as fighter jets and also have multiple, professional pilots at the controls who have been trained in IFR flight.
“Controlled flight into terrain” is a major cause of crashes, but these happen when the pilot thinks he knows what’s going on; a pilot in this situation won’t use the button.
The only case I can think of where PARS might have been helpful was Aeroflot 593 (mentioned upthread) - and even then they weren’t disoriented so much as actively overcorrecting in an attempt to restore level flight. Had there been PARS on their plane, I’m not sure they would have elected to use it.
Can anyone think of any others?