二目标空间中的一种并行分解型多目标进化算法

he multiobjective evolutionary algorithm based on decomposition (MOEA/D) borrows the decomposition idea from traditional mathematical programming methods. MOEA/D has shown remarkable performances when solving most multiobjective optimization problems (MOPs). However, MOEA/D still suffers from a very long running time for MOPs with large problem sizes or expensive objective evaluations. In this paper, a parallel MOEA/D (pMOEA/D) using the hybrid MPI and OpenMP programming model is proposed for bi-objective optimization. The population in pMOEA/D is distributed among computing nodes. Each computing node runs an MPI process for inter-node communications while each processor core runs an OpenMP thread to execute the iterations assigned for its resident node. Experimental results on a multi-core cluster with four nodes show that the parallel algorithm achieves significant performances in terms of speedup.