A Decomposition-based Constrained Multi-objective Evolutionary Algorithm with Adaptive Weight Adjustment

Yuki Azuma, Naoki Masuyama, Yusuke Nojima

Abstract: Many real-world optimization problems have multi-ple conflicting objectives and constraints to be considered simultaneously. These problems are referred to as constrained multi-objective optimization problems (CMOPs). To handle CMOPs, constrained multi-objective evolutionary algorithms (CMOEAs) have been proposed so far, including CMOEA with problem transformation from a CMOP to two-objective problems (CM2T). CM2T decomposes a CMOP into a set of two-objective subprob-lems using uniformly distributed weight vectors. In general, the uniformity of weight vectors can improve the solution diversity along the Pareto front (PF). However, it cannot work well when the PF is disconnected or is composed of feasible and infeasible regions. This paper incorporates an adaptive weight vector adjustment mechanism into CM2T (CM2T-AWA). We compare CM2T-AWA with six state-of-the-art CMOEAs on two constrained test suites. The experimental results shows that the proposed algorithm is superior to the comparison algorithms.