[scottm87]

You would be suprised how much the CMUCam already has integrated into it. Calibration routines are automated, and there is a fair amount of documentation on the site on how to deal with different lighting conditions. The camera routines that would are dependent upon the lighting conditions are integrated into the camera for the most part anyways.

In short, I think that the lighting will be less of a problem than everyone is making it out to be. Because so much of the camera interface is high level results of very complex, optimized, and tested machine vision routines, I think you will find the output that actually arrives to your robot is pretty clean as long as the lighting is fairly even and bright enough to illuminate the field.

I haven't tested it though with the IFI routines, so I don't know. I have, though, seen a CMUCam in action and was very impressed.