应用于大规模SoC软硬件划分的多种群协同进化方法研究  

Multi-group co-evolutionary method research:applied in large scale SoC hardware/software partitioning

作　　者：焦素敏[1] 王彩红[1] 王学梅[1] 

机构地区：[1]河南工业大学信息科学与工程学院,郑州450001

出　　处：《计算机应用研究》2011年第10期3690-3693,共4页Application Research of Computers

摘　　要：针对非支配遗传算法NSGA-Ⅱ在大规模SoC软硬件划分中存在的不足,鉴于协同进化机制在解决复杂问题上所体现出来的优势,提出了一种基于协同进化的混合算法CO-NSGA-Ⅱ,并引入了三个协同算子:种间交叉算子、淘汰算子和分裂算子。实验表明,在基准测试函数实验和较大规模的软硬件划分实验中,CO-NSGA-Ⅱ的性能要明显优于NSGA-Ⅱ,这也体现了协同进化在解决大规模软硬件划分问题上的优势和潜力。Aiming at the shortage of NSGA-II in large scale SoC hardware/software partitioning,and inspired by the advantage of co-evolutionary＇ performance on complex problems,this paper proposed a hybrid algorithm CO-NSGA-II based on co-evolution which had three co-evolutionary operators：interspecific hybridization,wash out and divide.Experiments show that,in benchmark test experiments and large scale hardware/software partition problems test experiments,CO-NSGA-II perform significantly better than NSGA-II.This result shows the advantage and the potential of co-evolution methods on solving large scale hardware/software partition problems.

关 键 词：协同进化 多目标优化 SOC软硬件划分 非支配遗传算法 协同进化非支配遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]