A parallel multi-objective cooperative co-evolutionary algorithm with changing variables

Abstract

Multi-objective optimization problems with changing variables are very common in real-world applications. This kind of problems often has a changing Pareto-optimal set and a complex relation among decision variables. In order to rapidly track the time-dependent Pareto-optimal front, we propose a framework of parallel cooperative co-evolution based on dynamically grouping decision variables. Decision variables are first divided into a number of groups using the Spearman rank correlation analysis, with different groups having a weak correlation. Then, a sub-population is employed to optimize decision variables in each group using a traditional multi-objective evolutionary algorithm. The evaluation of a complete solution is fulfilled through the cooperation among sub-populations. We compare the proposed algorithm with three state-of-the-art algorithms by applying them to two modified benchmark optimization problems. Empirical results show that the proposed algorithm is superior to the compared ones.