Dynamic constrained multi-objective evolutionary algorithms with a novel selection strategy for constrained optimization

The recently proposed dynamic constrained multi-objective evolutionary algorithm (DCMOEA) is effective to handle constrained optimization problems (COPs). However, one drawback of DCMOEA is it mainly searches the global optimum from infeasible regions, which may result in the bias against feasible solutions. Then the useful information about the optimal direction of feasible regions is not fully used. To overcome this defect, this paper proposes a novel selection strategy based on DCMOEA framework, called NSDCMOEA to solve COPs. The performance of NSDCMOEA is evaluated using a set of benchmark suites. Experimental results validate that the proposed method is better than or very competitive to five state-of-the-art algorithms.