A constrained multi-objective evolutionary algorithm using valuable infeasible solutions

Abstract

Many studies have shown that the effective mixture of feasible and infeasible solutions is beneficial to solve the constrained multi-objective optimization problems (CMOPs). So, it is of great importance to fully identify the valuable infeasible solutions that are helpful to find the constrained Pareto fronts. To achieve this, this paper evaluates the potential value of each infeasible solution from different aspects, and forms a criterion to fully identify the valuable infeasible solutions. Comparative analysis shows that compared with the existing constraint handling techniques, the proposed criterion is more effective in identifying valuable infeasible solutions. Accordingly, we embed the proposed criterion in the evolutionary algorithm and design a criterion-based constrained multi-objective algorithm for CMOPs. In the proposed algorithm, an archive preferring feasible solutions is used to record the best solutions, which are then mixed with the valuable infeasible solutions identified by the criterion to deal with the CMOPs. A series of simulation experiments on several benchmarks and engineering problem show the competitive performance of the proposed algorithm. Compared with the other state-of-the-art constrained evolutionary multi-objective optimization algorithms, the proposed algorithm performs better in dealing with different types of CMOPs, and it is the only one that can successfully deal with the problems that the initial population arises in the complex infeasible regions below the Pareto fronts.