嵌入式系统动态数据结构优化的并行进化算法

Parallel evolutionary algorithm for dynamic data structures optimization in embedded system

为了更好地解决现代多媒体嵌入式系统动态数据结构优化问题,结合NSGA-II和SPEA2两个多目标进化算法,引入岛屿模型和多线程机制,提出了一种并行多目标进化算法——PMOEA-NS。基于多核计算机系统,使用PMOEA-NS具体的3个不同并行算法和串行NSGA-II、SPEA2,对一个实际动态嵌入式应用程序进行优化实验和计算,结果表明:与串行算法NSGA-II和SPEA2相比,并行算法不但提高了优化过程的速度,而且改善了解的质量和多样性。

In order to better solve dynamic data structures optimization in embedded system, this paper combined NSGA-II and SPEA2, and adopted island model and multi-thread technique to describe a parallel multi-objective evolutionary algorithm. Using its specific three parallel algorithms and sequential NSGA-II and SPEA2, one embedded application on multi-core architecture was optimized in experiment. The results show that not only the speed of optimization process is enhanced, but also the quality and the variety of the solutions was improved.