一种用于任务调度的广义遗传算法

Generalized Genetic Algorithm for Task Scheduling

提出一种用于解决网格系统并行计算中任务调度问题的遗传算法。通过设计一维十进制分离编码的特殊编码方式,使分配子串和调度子串的杂交与变异分离进行。借鉴广义遗传算法的进化思想,使种群中的母本在进行交叉和变异后也参与下一代的选择竞争。模拟实验结果证明,该算法与二维矩阵编码和精英选择策略相比,能得到更好的解且收敛速度更快。

This paper proposes a genetic algorithm for task scheduling in grid parallel computing. By designing a special coding method that one coding string contains two substrings, and each substring has different cross and variation methods. An evolutionary process of generalized genetic algorithm is adopted which makes the mother competes with the children also. Simulated experimental results show that the algorithm outperforms the algorithms using 2D coding or elite selecting strategy, it can gain better results and its convergent speed is faster.