约束多目标进化算法修补算子的研究

Research on Repair Operators in Constrained Multi-objective Evolutionary Algorithm

为了避免约束多目标进化算法陷入局部最优,提出了一种新的边界修补算子.该边界修复算子受到反向学习的启发,把违法盒型约束的解修复到其对应的反向可行边界,以增强约束多目标进化算法的多样性.为了验证所提的修补算子的有效性,在经典的约束多目标基准测试问题CTP2-CTP8上进行了实验仿真,仿真的结果表明所提出的新型的修补算子在多样性和收敛性上均优于现有的边界修补算子.为了进一步验证所提出的新型修补算子,设计了一组约束多目标优化问题MCOP1-MCOP7,作为CTP测试问题的有效补充.在MCOP1-MCOP7上的仿真结果同样表明,所提出的新型边界修补算子同时在收敛性和多样性上要优于现有的修补算子.

In order to avoid falling into local optimum for constrained multi-objective evolutionary algorithm, we design a new repair operator which employs a reversed correction strategy to fix the solutions that violate the boxconstraint. This repair operator inspired by the concept of opposition-based learning. It fixes the infeasible solution that violates the box-constraint to its reversed feasible boundary, so that it can help to increase the diversity of constrained multi-objective evolutionary algorithm. We test the proposed repair operators and other existing repair operators in the framework of MOEA/D on CTP2 to CTP8 instances, the experimental results validate the proposed repair operator is better than existing repair operators in terms of both convergence and diversity. To further demonstrate the performance of proposed repair operator, we design a set of multi-objective constrained optimization problems named MCOP1 to MCOP7, as a complement of CTP benchmark test problems. The test results on MCOP1 to MCOP7 also show that the proposed repair operator is better than existing repair operators.