[gpm]Tracking fast moving objects with optical mice.

[Christian Melki]

Hi, thought that this might be the place to turn for help.
Im currently building robots to participate in robotic
soccer this summer. We have some ideas of tracking
our position with diffrent methods, like ultrasonic arrays
and neural decision making. Some of them are implemented.
and some are not.. Like this idea I have of using optical mice
of tracking the ground and recieving dx,dy coords.
I have written a small program that reads the mouse fd
and outputs the dx, dy coords acoording to our needs..
But after small scale testing I have encounterd several
problems.

1. The mouse doesn't seem to track the ground fast enough.
If i move it _slowly_ across a surface i can get it to accurately 
give me a dx-dy in metric format after scaling with resolution.
But if i move it more a bit faster.. it seems to loose track of
whats happening.. It recognizes a change.. and returns a 
value accordingly.. But theese are close to a dx-dy of 0, 0
instead of beeing a high value as an indication of a fast 
motion. I have tested it with a budget optical  mouse as 
well as an more expensive microsoft imps2 type.. 
The results are the same.. I cant move nearly as fast as 
I want.. Although the mice reportedly uses a 5000 
samples/sec mechanism. I cant remember needing to move 
this slowly on my mousepad when playing games.. Are
there any ps2 specific options that im missing? Setting
mouserate.. setting resolution..? Are there any imps2
specific options to increase tracking abilities at high speeds?

2. The init seems to be broken on most occations when
initing the mouse.. More than half of the coordinates go
"beserk", reporting 0-255  type movements.. And on a few
occations everything seems ok.. All values are ok, no matter
what surface or how long i run the mouse.. Seems very strange
to me.. Any ideas of what i might be doing wrong?
I have looked a bit at the gpm code.. albeit a bit to big and
complex for me.. I have also looked at the svgalib mouse
implementation.. 

In short.. I could use any tips and pointers I can get. :)
Best regards.

Christian Melki
MScE IT Stud.
Uppsala University
Uppsala, Sweden.