基于多目标遗传算法的模拟电路进化设计方法

提出一种面向电路进化设计的多目标自适应遗传算法，利用均匀设计技术合成多个适应度函数以提高搜索方向的空间均匀性，利用基于元件标称值的网表形式高效编码方案来支持电路结构自动生成和提高设计结果的实用性，利用基于均匀设计的多个体交叉算子来提高交叉操作的效率和采样均匀性，利用跟随遗传进程并区别不同基因位的遗传概率调整策略来提高进化效率和全局收敛率．实验结果表明，该方法可用较小的运算量获得符合设计目标的多种设计结果．

A uniform design based multi-objective adaptive genetic algorithm towards the evolutionary design of analog circuits

The Multi-Objective Adaptive Genetic Algorithm(MOAGA) proposed in this paper is aimed at the evolutionary design of circuits. It features a group of uniformly scattered search directions toward the Pareto frontier obtained by using the Uniform Design Technique(UDT), an efficient circuit representation scheme that enables circuit structures to be generated and improved automatically to get practical design results, a UDT based multi-parents multi-offspring crossover operator to improve the quality of population with a less computational cost, and an adaptation technique that enables genetic parameters P_c and P_m to track the evolution process and treat each type of loci respectively according to its effect on the fitness so as to increase search efficiency and global convergence possibility. Some experimental results presented show that the MOAGA is capable of searching out effective circuits in accordance with the required responses, while the computational cost can be largely decreased.