Those are complex issues about disengaging autopilot.  Your comments
helped me to understand why the FG aircraft sometimes does gyrations when disengaging
the VOR NAV setting Ctrl-N.  I'll try setting the controls to neutral before hitting Ctrl-N
to disengage.
 
Tom
 
 
----- Original Message -----
From: John Denker
To: FlightGear developers discussions
Sent: Saturday, June 30, 2007 12:15 PM
Subject: Re: [Flightgear-devel] [off list] patch forcontrol lockingbyautopilot

On 06/30/2007 09:52 AM, tmpargolf@bellsouth.net wrote:
> An airliner some years ago crashed into the Everglades in Florida
> because the autopilot
> was unknowingly disengaged by accidental knee pressure on the Yoke as
> the pilot
> was getting out of his seat.  Specs showed that the minimum 45 pounds
> pressure required was faulty.

> The aircraft had been in a holding pattern pending confirmation that the
> gear was down, since the
> indicator lamp was burned out and the spare broke upon attempted removal
> from its recessed
> location.  Unknowingly the aircraft was slowly descending and the last
> thing on the voice
> recorder ws the co-pilot "Hey, there's something wrong with the
> instruments" when he
> noticed the altimeter showed just above ground level.

> If I understand the FG issue correctly, then I would think that a sudden
> movement of
> the Yoke could be used to disengage the autopilot.

This is a tricky issue.

One case that has to be considered is what happens when you
are /engaging/ the autopilot.  In particular, suppose you
are in a long-winged glider in a steep turn, holding
tons of outside aileron to compensate for the overbanking
tendency.  You engage the autopilot, desiring that it
will maintain the turn.  Then
  -- in the real aircraft, you let go of the yoke and
   it stays where it is.  No problem.
  -- in FG, you let go of the yoke and it springs back
   to the center.  That's a problem.

On the other side of the same coin, suppose you want to
disengage the autopilot while the ailerons are deflected.
You really ought to deflect the joystick so that it matches
what the autopilot is doing with the yoke, before hitting
the autopilot disengage button.

Similar considerations apply to the pitch axis and yaw axis.
Keep in mind the pilot who inadvertently snap-rolled a 747,
seriously injuring a couple of passengers, by disengaging the
autopilot without noticing that the autopilot was holding one
of the rudder pedals to the floor.
   http://en.wikipedia.org/wiki/China_Airlines_Flight_006
   and references therein.


Note that there are four stages of sophistication among FG users:
  a) Using the keyboard for primary flight control;
  b) Using the mouse for primary flight control;
  c) Using a plain old joystick for primary flight control; or
  d) Using a joystick with force-feedback (or position servos)
   for primary flight control.

It is slightly peculiar that the problem is only serious in
stage (c).  It does not arise in stage (b) because we can warp
the mouse to the desired position;  we let the user deal with
the side effects of such warpage.

Arguably the theoretically-ideal solution would be for everybody
to skip stage (c) and go directly to fully position-servoed
joysticks ... but that is not likely to happen anytime soon,
so for now we are still facing nontrivial problems at stage (c).

Note that the problems are compounded by the fact that the naive
user does not know what to expect ... and indeed doesn't even
understand what he's seeing when a war breaks out between the
autopilot and the joystick.  It just looks like something is
broken.  Disabling the joystick when the autopilot is active
ends the war, but doesn't really solve the user's problem;  he
just sees it as a different kind of brokenness.

One half-baked idea I've been toying with involves animating a
/hand/ which is normally gripping the yoke.  The joystick moves
the hand.  When the autopilot is engaged, the joystick still moves
the hand, but the hand is not gripping the yoke.  I'm not sure
how hard this would be to implement.  In any case, it has some
conceptual value, providing a way to visualize the nature of the
problem, to some extent.

This is an important topic to be discussing.  Some of the recent
suggestions are commendable steps in the right direction, but I
reckon we are still one breakthrough removed from a complete
solution to the problem.

-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/
_______________________________________________
Flightgear-devel mailing list
Flightgear-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/flightgear-devel