As mentioned in the text, this approach uses non-Gaussian mutation.  
It is my belief now that the improvement in performance seen in the 
CAM approach is likely attributable to the fact that the selected 
mutation method biases extreme variable values and that the test 
problems all feature optimal points on or about the extreme values 
of the search space.  Further experiments seem to have borne this 
out and I am of the belief that the approach is inferior to NSGA-II, 
despite reported findings.

Adam Berry
24 June 2011