基于Memetic算法的电路演化设计研究

Research on Evolvable Hardware Design Based on Memetic Algorithm

针对传统演化算法在设计数字逻辑电路时存在的演化速度缓慢和容易陷入局部最优解等问题,设计了一种Cartesian进化编程编码的电路演化Memetic算法,采用遗传算法作为全局搜索方法,并设计了适合电路演化的基本门种类局部搜索策略.通过一位全加器电路证明所提出的Memetic硬件演化算法的搜索能力,实验证明所提出的算法能够有效地完成进化任务,具有较强的全局快速和局部搜索能力.

In order to overcome the problems of traditional evolution algorithms in designing digital logic circuits,such as slow evolution speed and premature convergence,a memetic algorithm（MA） for evolving digital circuits is presented based on CGP encoding.It adopts genetic algorithm as global search method.A local search strategy based on gate type is proposed.The search ability of the proposed method is tested by evolving one-bit full adder.Experimental results show that it is efficient in solving the problems of hardware evolution and has the ability of fast global and local convergence.