Abstract: Aiming at the software/hardware partitioning problem of the embedded system, this paper presents a modified Differential Evolution(DE) algorithm by employing the random mutation strategy on the foundation of hybrid DE algorithm. The modified algorithm extends the influence of the best vector and meanwhile mitigates the impact upon mutation vectors brought about by the best vector. Experimental results demonstrate that compared with the hybrid DE algorithm, the modified algorithm enjoys faster convergence speed and higher reliability, and can effectively solve the problem of hardware-software partitioning.

Key words: stochastic optimization algorithm, constrained optimization, Differential Evolution(DE) algorithm, mutation strategy, embedded system, software/hardware partitioning

摘要： 针对嵌入式系统的软硬件划分问题，在混合优化策略微分进化算法的基础上，提出一种采用随机变异策略的改进微分进化算法。在扩大最优向量影响范围的同时，减小变异向量受最优向量影响的程度。实验结果表明，与混合优化策略微分进化算法相比，该算法具有更快的收敛速度和更高的可靠性，能够较好地解决嵌入式系统的软硬件划分问题。

关键词: 随机优化算法, 约束优化, 微分进化算法, 变异策略, 嵌入式系统, 软硬件划分

CLC Number: 

• TP311