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

Abstract: 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.

Key words: software project scheduling, dynamic software project scheduling, multi-objective evolutionary algorithm, two-archive evolutionary algorithm