Autopilot disconnection protocol
 

A paper suggests that a new protocol for sudden autopilot disconnection (as opposed to e.g. more pilot training) might have averted some high-profile accidents.

Redesign for sudden autopilot disconnection needed, say flight safety experts

(Disclaimer: I'm not a pilot.)

An excellent idea. The original protocol comes from a long distant time when pilots were pilots and could be counted on to fly the aircraft in situations where the autopilot was beyond its limits.

Now we're often hoping that the autopilot will take care of the plane when the pilots get beyond their limits.

A system where the autopilot continues to try flying the plane while warning the pilots that it is beyond limits is definitely overdue.

The ‘paper’ – as assessed on half a web page of weak reporting, is unworthy of a science title.
The safety experts ‘findings’ (when/if published) might only reflect common knowledge which represents the symptoms of part of an event; they are not a cause.
It’s human nature to be surprised, the degree of which depends on the situation, expectedness (training), or personal qualities in managing emotion. This can increase workload, but that is what a pilot should be trained to manage.

Certification requirements (CS / FAR 25) require that sudden autopilot disconnects (required to avoid serious safety hazards) must not result in the aircraft being significantly out of trim. In all of the accidents quoted, the aircraft deviations subsequent to AP disconnect resulted from crew activity; an activity presumably triggered by thought or assessment of the situation. The importance of these situational aspects is not an issue of systems’ activity – why did the autos fail (its history), but what is the aircraft doing, and what do ‘I’ wish the aircraft to do (proactive awareness).

‘Autos should transfer control with the same protocol as pilots’!!!
If only pilots would transfer control with the same provisos as current certification … , then there would be fewer dusted knuckles for those who have not trimmed the aircraft beforehand.
Whereas a pilot can communicate difficulty in flying or increasing workload, most autos cannot. Seeking for ‘automation context’ is like mind reading before a system failure.
This is crystal-ball stuff, not science; and as for the origin (hfes), perhaps they should refresh themselves of what their mission states; ‘… compatibility in the design of interactive systems of people, machines, and environments…’
Not man or machine, but both within the environment; and all considered as an interactive system.

If, for some reason the autopilot fails to do what the pilot expects it to do, or disengages without good reason, then the switch-over by the pilot to hand flying should be seamless and without raising a sweat.

Mostly that doesn't happen because an autopilot unexpected disengagement is usually followed by frantic attempts at re-engagement followed by heads down button pushing. Addiction to automatics is often almost incurable. It can be likened to addiction to nicotene. Both addictions have the potential to cause eventual grief.

I hope this is severe journalistic misinterpretation. If it is indeed so written in the paper then the folks trying to improve autopilot have no clue how it works. It is designed to disconnect itself promptly when there is slightest doubt about its ability to keep aeroplane upright so it doesn't get it into unrecoverable attitude.

So it is supposed to be clairvoyant?!?!? Like: I am fine for the time being but your pitots might get clogged int his cloud ahead of us so I'm warning you I'll pack up if airspeed gets unreliable. Actually it would be even better if we could replace those scary red flashing lights and cavalry charges with HAL 9000 voice; "Dave.... I'm about to turn myself off."