This paper improves upon the reference NSGA-II procedure
by removing an instability in its crowding distance operator.
This instability stems from the cases where two or more indi-
viduals on a Pareto front share identical _tnesses. In those
cases, the instability causes their crowding distance to ei-
ther become null, or to depend on the individual's position
within the Pareto front sequence. Experiments conducted
on nine di_erent benchmark problems show that, by com-
puting the crowding distance on unique _tnesses instead of
individuals, both the convergence and diversity of NSGA-II
can be signi_cantly improved.
Categories and Subject Descriptors
I.2.8 [Arti_cial Intelligence]: Problem Solving, Control
Methods, and Search|heuristic methods