基于改进双归档进化算法的多目标动态软件项目调度  

Multi-objective dynamic software project scheduling based on improved two-archive evolutionary algorithm

作　　者：陈志远 伍章俊[1,2] 童珊珊 刘晓 CHEN Zhiyuan;WU Zhangjun;TONG Shanshan;LIU Xiao(School of Management, Hefei University of Technology,Hefei 230009, China;Key Laboratory of Process Optimization and Intelligent Decision-making, Ministry of Education, Hefei University of Technology, Hefei 230009, China;School of Information Technology, Deakin University, Geelong VIC 3126, Australia)

机构地区：[1]合肥工业大学管理学院,安徽合肥230009 [2]合肥工业大学过程优化与智能决策教育部重点实验室,安徽合肥230009 [3]迪肯大学信息技术学院,澳大利亚吉朗VIC3126

出　　处：《计算机集成制造系统》2021年第9期2565-2574,共10页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(71871076);安徽省自然科学基金资助项目(1708085MG169)。

摘　　要：项目需求变化和开发人员流动使得软件项目管理具有动态性的特征。鉴于此,建立了包含持续时间、项目成本、调度鲁棒性和调度稳定性的多目标动态软件项目调度模型,并提出一种改进的双归档进化算法。双归档进化算法在优化多目标约束问题时可以同时平衡收敛性、多样性和可行性。但随着目标维度增加,双归档进化算法的性能会下降。本文提出的改进算法采用佳点集和启发式策略进行种群初始化,利用评价函数自适应地对两种交叉和变异方法进行概率选择,分别采用质量指标和动态拥挤度距离对收敛性档案和多样性档案进行更新。对比实验基于仿真的和真实的软件项目进行。结果表明,改进的双归档进化算法具有良好的性能,可以获得质量更高的帕累托解集。The software project management is dynamic caused by variation of project requirements and flow of developers.A multi-objective dynamic software project scheduling model including duration,project cost,scheduling robustness and scheduling stability was established.The two-archive evolutionary algorithm could balance convergence,diversity and feasibility when optimizing the multi-objective constraint problem.However,the performance of the two-archive evolutionary algorithm was decreased with objective dimension increasing.To address such a problem,the optimal point set and heuristic strategy were used for population initialization.The evaluation function was used to adaptively determine the probability of two methods of crossover and variation.Quality index and dynamic crowding distance were used to update convergence archive and diversity archive.Based on simulation and real software project,the experimental results showed that the improved two-archive evolutionary algorithm had good performance and could obtain higher quality Pareto solution sets.

关 键 词：软件项目调度 动态软件项目调度 多目标进化算法 双归档进化算法 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论]