A constrained multiobjective evolutionary algorithm based decomposition and temporary register

We propose a novel constrained multiobjective evolutionary algorithm based on decomposition and temporary register in this paper. It decomposes the constrained multiobjective optimization problem into a number of subproblems and then optimizes each subproblem in a collaborative way. We also propose a novel constraint handling technique based on temporary register. Each subproblem has its own subpopulation and one temporary register. The subpopulation is composed of those individuals which has better objective values and lower constraint violations of this subproblem, while the temporary register is composed of those individuals that are found before. We perform the crossover operator between each individual in the subpopulations and an individual which is randomly chosen from the corresponding temporary register. Therefore, the temporary register strategy makes the individuals which have better objective values and lower constraint violations have an opportunity to participate in the crossover and mutation, but don´t been eliminated at once. Moreover, this constraint handling technique does not need any parameter setting. The numerical simulations show the proposed algorithm outperforms existing ones.