A Novel Two-Stage Evolutionary Algorithm for Constrained Multiobjective Optimization

When tackling constrained multi-objective optimization problems (CMOPs), especially problems with complex feasible areas, it is challenging for handling both objective optimization and constraint satisfaction. To remedy this problem, this paper introduces a novel two-stage constrained multi-objective evolutionary algorithm, NTEA, with different emphases on the objectives and constraints. In the first stage of NTEA, a dynamic balance method is used to dynamically adjust the selection preference from Pareto dominance to constrained dominance. The purpose of this stage is to obtain a certain feasible solutions and solutions with good objective values, thereby preventing the population from becoming trapped in local optima. In the second stage, an efficient search method based on competitive swarm optimizer is used to accelerate the convergence to the constrained Pareto front while maintaining the diversity of the whole population. To test the effectiveness of the proposed NTEA, experiments are carried out on two popular benchmark suites LIR-CMOP and DAS-CMOP. The outcomes demonstrate that the suggested algorithm competes effectively with state-of-the-art constrained multi-objective optimization evolutionary algorithms (CMOEAs).