A Two-Population Evolutionary Algorithm for Constrained Optimization Problems

A new approach to solving constrained nonlinear programming problems using evolutionary computations is discussed. According to the method two populations are evolved, one population (females) is evolved inside the feasible domain of the design space and a second population (males) is evolved outside this feasible domain. Both populations can be independently subject to crossover and mutation operations and the design space explored. Female-male crossover however ensures the desirable increase in the search pressure upon the boundaries of the feasible space -it is known that in many optimization problems the global optimum is bounded. The experiments performed on three test objective functions of two variables show some promise of the proposed approach in that it can cope with both linear and nonlinear constraints and with nonconvex feasible domains.