GA for Quadratic Assignment Problem (QAP)

i am trying to use matlab ga optimtool to derive a solution for QAP. in this problem, the feasible solution can be either 1 of the following cases:

1. a permutation of n. e.g. permutation of 3 - 2 1 3

2. a permutation matrix

i manage to write the fitness function, however, i just cant set the population to always be feasible solution. in other words, after a few iterations, the chromosome becomes something like (0.21 -354 1) rather than (1 3 2)

anyone have any idea to model this problem using matlab ga tool?

thanks in advanced!