时空视角下的动态多目标进化算法研究综述

现实中的多目标优化问题会随着时间或环境的变化而发生改变,因此在全周期优化过程中,环境变化检测和算法响应是求解动态多目标优化问题的两大关键步骤,为此重点对动态多目标进化算法方面的研究进行总结.为有效求解动态多目标优化问题,大量追踪性能优良的动态多目标进化算法在近20年里被提出,但是很少有文献从时空角度对已有研究进行分析和报道,鉴于此,从该视角对动态多目标进化算法研究进行综述.首先介绍动态多目标优化的基本概念、问题和性能指标;然后从时空视角对近5年提出的动态多目标进化算法研究进行分别介绍;最后列出目前动态多目标进化算法方面研究存在的一些挑战,并对未来研究进行展望.

A research survey of dynamic multi-objective evolutionary algorithms from spatiotemporal perspective

Actual multi-objective optimization problems change with time or environments(called as dynamic multi-objective optimization problems, DMOPs), thus detection of environmental change and algorithm response are two key steps to solve DMOPs during the full-cycle optimization. This paper focuses on summarizing the research on the latter one, i.e., dynamic multi-objective evolutionary algorithms(DMOEAs). To solve DMOPs effectively, a large number of DMOEAs with good tracking performance have been proposed in the past two decades. However, few literatures analyse and report the existing studies from the spatiotemporal perspective. Therefore, this paper introduces review of research on DMOEAs from this view. First, the basic concepts, DMOPs, and performance indicators are introduced. Then, the research on DMOEAs proposed in the past five years are introduced from the spatiotemporal view. Finally, some current challenges exist in DMOEAs are given, and future studies are prospected.