摘要
针对软件和硬件实现方式各自的优点及不足,提出了遗传算法的软硬件协同设计方法,并且将这种方法在FPGA上进行了具体的实现。首先对遗传算法流程中的各个模块进行了详细的分析,根据软硬件的不同特点以及设计实现的目标,对遗传算法的功能模块进行了软硬件划分,然后对硬件实现的部分进行了详细的介绍,包括模块之间的连接,模块的内部状态机,模块的端口,所有硬件模块的功能仿真。同时,为了保证软硬件之间的正常通讯,提出了一种新的软硬件交互通讯协议。最后将硬件实现部分做成通用IP核,方便其他设计者使用,并给出了软硬件协同的遗传算法在二进制问题和0-1背包问题两个实例中的具体应用数据。通过与纯软件实现方式的实验数据进行比较,提升了算法运行时间50%的效率,且算法的收敛性保持一致,进一步验证了该算法的适用性及高效性。
In view of the advantages and disadvantages of software and hardware implementation,a software hardware co-design method based on genetic algorithm is proposed and implemented on FPGA.We analyze each module of the genetic algorithm flow chart,divide the function module of genetic algorithm in hardware and software according to their different features and design purpose.After that,we elaborate the implementation of hardware part,including connection between modules,state machine in a module and interfaces of a module,function simulation of all hardware modules.We also provide a protocol to guarantee the communication between hardware and software.Ultimately,the hardware implementation is designed as common IP kernel to be reused by other designers.Then we give the application data of software hardware cooperation genetic algorithm in binary problem and 0-1 knapsack problem.Compared with the experimental data of pure software implementation mode,the efficiency of the algorithm running time is improved by 50%,and the convergence of the algorithm is consistent,which further verifies the applicability and high efficiency of the algorithm.
作者
聂鑫
殷若兰
刘海峰
NIE Xin;YIN Ruo-lan;LIU Hai-feng(Hubei Key Laboratory of Intelligent Robot,Wuhan Institute of Technology,Wuhan 430205,China;School of Computer Science and Engineering,Wuhan Institute of Technology,Wuhan 430205,China;Huawei Technologies Co.,Ltd.,Shenzhen 518000,China)
出处
《计算机技术与发展》
2021年第11期114-121,共8页
Computer Technology and Development
基金
国家自然科学基金资助项目(61672391)
智能机器人湖北省重点实验室引导基金资助项目(HBIRL202009)。
关键词
遗传算法
FPGA
软硬件协同
软硬件划分
IP核
0-1背包问题
genetic algorithm
FPGA
software and hardware cooperation
software and hardware division
IP core
0-1 knapsack problem
作者简介
聂鑫(1983-),男,博士,副教授,硕导,CCF会员(E0047M),研究方向为演化算法、演化硬件;殷若兰(1997-),女,硕士研究生,研究方向为软件工程。
