基于决策变量关系的动态多目标优化算法

动态多目标优化问题(DMOPs)需要进化算法跟踪不断变化的Pareto最优前沿,从而在检测到环境变化时能够及时有效地做出响应.为了解决上述问题,提出一种基于决策变量关系的动态多目标优化算法.首先,通过决策变量对收敛性和多样性贡献大小的检测机制将决策变量分为收敛性相关决策变量(CV)和多样性相关决策变量(DV),对不同类型决策变量采用不同的优化策略;其次,提出一种局部搜索多样性维护机制,使个体在Pareto前沿分布更加均匀;最后,对两部分产生的组合个体进行非支配排序构成新环境下的种群.为了验证DVR的性能,将DVR与3种动态多目标优化算法在15个基准测试问题上进行比较,实验结果表明, DVR算法相较于其他3种算法表现出更优的收敛性和多样性.

A dynamic multi-objective optimization algorithm based on the relationship of decision variables

Dynamic multi-objective optimization problems require evolutionary algorithms(EAs) to track the changing Pareto-optimal front(PF) at different times, then can respond effectively and timely when environmental changes are detected. In order to solve the above problem, a dynamic multi-objective optimization algorithm based on the relationship of decision variables is proposed. Firstly, through the detection mechanism of the contribution of decision variables to convergence and diversity, the decision variables are divided into convergence decision variables(CV) and diversity decision variables(DV). Secondly, different optimization strategies are adopted for different types of decision variables. And a local search diversity maintenance mechanism is proposed to make individuals more evenly distributed in the Pareto front. Finally, the non-dominated sort of the combined solutions generated by the two parts constitutes the population in the new environment. In order to verify the performance of relationship of decision variables, relationship of decision variables is compared with three dynamic multi-objective optimization evolutionary algorithms on the 15 benchmark functions. Experimental results demonstrate that the DVR algorithm exhibits better convergence and distribution than the other three algorithms.