混合NSGAII和DE的优化算法及应用
2018-12-21李岩张光武
李岩 张光武
摘要:遗传算法NSGAII在引入快速非支配排序算法、拥挤度算子以及精英策略后重复个体产生的概率明显上升,降低了帕累托效率。针对这一缺陷进行了改进,去除了重复个体并保持种群数量不变。根据遗传算法基因交叉变异的方法和差分进化算法DE的思想,将改进后的NSGAII算法与DE算法进行有效混合构建一种新的多目标优化算法。通过MATLAB对优化后的算法进行验证,结果表明优化后的算法在分布性和收敛性上都有所提高,搜索解的能力也有所提升。然后利用优化后的算法完成对μC/OSII任务管理部分的软硬件划分。
关键词:遗传算法;NSGAII;差分进化算法;MATLAB;软硬件划分
DOI:10.15938/j.jhust.2018.05.013
中图分类号: TP3162
文献标志码: A
文章编号: 1007-2683(2018)05-0075-05
Optimization Algorithm and Application of Hybrid NSGAII and DE
LI Yan,ZHANG Guangwu
(School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China)
Abstract:Due to the introduction of fast nondominated sorting algorithms, crowding operators and elite strategy, the probability of repetition individual increased significantly in every population of NSGAII algorithm, reducing the Pareto efficiency It has been improved for this defect, removed the repeating individual and maintained the number of populations unchanged According to the genetic algorithm crossover and mutation method and differential evolution algorithm DE, the improved NSGAII and DE are combined to construct a new multiobjective optimization algorithm The algorithm takes DE as the main optimization method, and uses the basic idea and crossover and mutation method of genetic algorithm The optimization algorithm was verified by MATLAB The results show that the optimized algorithm has been improved in both distribution and convergence, and the capacity of search solution has also been improved At last , the optimization algorithm is used to complete the hardwaresoftware partitioning of task management part in μC/OSII
Keywords:genetic algorithm; NSGAII; differential evolution algorithm; MATLAB; softwarehardware partitioning
0引言
遺传算法[1]的研究始于20世纪50年代,在80年代末有了快速发展,并在人工智能和控制系统等领域得到了广泛的应用。在应用中也逐渐暴露了遗传算法的缺点,研究者开始针对遗传算法在具体方向上的缺点进行一系列的改进[2-6]。其中,Deb在2000年对NSGA算法提出了改进,算法NSGAII降低了算法的复杂度。NSGAII算法在执行过程中,经过交叉、变异操作很容易在同代以及父代之间生成相同个体,算法搜索新解的能力也因此受到很大影响,这种情况将导致算法运行的结果很容易是局部最优,不仅使得解很可能会丢失而且也会对算法的收敛速度有影响。文章首先对算法产生重复个体的问题进行了改进,改进后的算法在分布性上较原算法有一定提高。
差分进化算法[7-10]是一种以群体智能理论为基础模拟生物进化的优化算法,简单且高效的将适应环境的个体保留下来。……
