Solving constrained optimization problem by a specific-design multiobjective genetic algorithm

By transforming the constrained optimization problem into a multiobjective optimization problem, a specific-designed multiobjective genetic algorithm is proposed. For this multiobjective optimization problem, the objectives transformed by constraints depend on the number of generations such that the algorithm initially makes the search in a region that can contain infeasible solutions and gradually concentrate the search in the feasible region. Therefore, the proposed algorithm is not sensitive to active constraints and can handle the constraints efficiently. In addition, a new kind of multiple fitness functions, defined by the maximum value of the normalized objective multiplied by weights, can aid the proposed algorithm to explore the search space uniformly, keep the diversity of the population, and distinguish the quality between the feasible solutions and infeasible solutions. The numerical simulations indicate the proposed algorithm is efficient.