摘要
提出一种利用可编程模拟器件AN231E04的粗粒度结构及高速重配置能力开发大规模模拟电路在线演化硬件的方法,并在此基础上设计ANEHP-Alpha在线演化平台.平台针对在线演化系统的特点,利用小规模贪婪搜索获取高适应度初始种群;以精英保留策略和控制参数动态调节避免种群早熟,确保算法收敛能力及速度;引入快速预评估器排除不良个体降低下载率,从而大幅提高演化速度并确保器件不受非法电路损害.此平台为高速大规模模拟电路在线演化方法研究奠定了较好的基础.
A new method is put forward to design large-scale intrinsic analog evolvable hardware (EHW) based on AN231E04, a new brand field programmable analog array (FPAA) device with big granularity and high reconfiguration speed. ANEHP-Alpha, an intrinsic EHW platform, is designed with this method. Small-scale greedy search method is employed to generate initial population with preponderant fitness value. Elite individual reservation strategy is used to ensure the ability of convergence and controlling parameters are dynamically adjusted to avoid earliness and increase the convergence speed. A fast pre-evaluation and bad individual elimination method is employed which highly increases the speed of evolution and ensures the devices against the damage induced by illegal individuals. ANEI-IP-Alpha lays a better foundation for the further research of large-scale intrinsic analog evolvable hardware.
出处
《武汉大学学报(工学版)》
CAS
CSCD
北大核心
2008年第2期116-120,共5页
Engineering Journal of Wuhan University
基金
国家“863”计划基金资助项目(编号:2002AA1Z1490,2007AA01Z290)
国家自然科学基金项目(编号:60442001,60773009)
关键词
演化硬件
FPAA
遗传算法
evolvable hardware (EHW)
field programmable analog array (FPAA)
genetic algorithm