Re: Ada / Boeing 777

[mbk@caffeine.engr.utk.edu (Matt Kennel)]

Sam Harbaugh (harbaugh@ACUSYS.COM) wrote:
: Tim Rowe <digitig@CIX.COMPULINK.CO.UK> wrote:
: >As an example, there was a known fault on one of the Airbus range (which
: >has been flying total fly-by-wire for *many* years before the 777!) that
: >meant that when flying a heading of 00 degrees, if the pilot commanded
: >the aircraft to turn a bit one way, the aircraft could actually turn the
: >other way. It was turning to the correct heading, but it was going the
: >359 degree route, not the 1 degree route. Now, there are not many runways
: >in the world that are close to north-south, but there are a few, and when
: >that bug showed itself on a final approach I bet the pilots pants turned
: >brown.

: Point of order:  heading zero is invalid in the user space.  north-south
: runways are marked 36 for 360 degrees, not zero.  Pilots and ATC speak of
: 360 degrees, not zero degrees.

: I wouldn't suggest that this point has anything to do with the malfunction
: but maybe, just maybe, if the programmer had a pilot's license he/she would
: have known this and then maybe, just maybe, he/she would have known to turn
: in the proper direction.

: Just maybe, just possibly maybe, if the code was written in a strongly typed
: language, using human readable names, and a pilot attended a code
: walkthrough, the pilot would have picked up on type heading_type being from
: 0..359 instead of 1..360 and started a conversation that would have lead to
: discovery of the worong way turn.

The representation (aka implementation) of angle is wrong.  

It should be a complex number of unit magnitude.  Rotations in any
direction can be implemented as complex multiplication without any
discontinuities, wrap around errors or problems at 0 or 180 or 360
degrees.   

This is an easy 2-d representation of the rotation group.  3-d is 
trickier: either Euler angles or quaternions. 

Don't just hack on the problem: transcend it. 

See, sometimes physicists *can* be useful.  ;-)