基于混沌遗传算法的模糊CMAC控制  

Fuzzy CMAC Control Based on Chaos and Genetic Algorithm

作　　者：张雷[1] 马建伟[1] 宋书中[1] 

机构地区：[1]河南科技大学电子信息工程学院,河南洛阳471003

出　　处：《计算机测量与控制》2012年第8期2117-2120,共4页Computer Measurement &Control

基　　金：河南省科技攻关计划项目(082102210015)

摘　　要：分析CMAC神经网络和模糊控制的特性,给出了一种能反映人脑认知的模糊性和连续性的模糊CMAC神经网络控制器,该控制器采用高斯函数作为模糊隶属函数,利用神经网络实现模糊推理并可对隶属函数进行实时调整,从而使其具有学习和自适应能力;针对BP算法易陷入局部极值点的缺点和简单遗传算法局部搜索能力差的不足,提出了一种混合学习算法,即首先利用混沌遗传算法全局搜索的特点来离线优化神经网络的参数,再利用BP算法较强的局部搜索能力对网络参数进行在线调整;仿真结果表明了该方法的可行性和有效性。Based on the characteristics of CMAC neural network and fuzzy control, the novel controller of fuzzy CMAC neural network that reflects the fuzziness and continuity of human cerebella is discussed, In the controller , Gauss function is used as fuzzy membership function, fuzzy inference is realized by CMAC neural network and the shape of member function can be adjusted in real time, which endures the controller with capability of learning and self--adapt. To counteract the defects of BP algorithm and the chances of simple genetic algo- rithm premature convergence, a hybrid learning algorithm was proposed. First, the chaos genetic algorithm was used to optimize the fuzzy neural network＇s parameters off--line. Then, because of the strong capability of local search, the BP algorithm was used to adjust the pa- rameters on--line. The simulation results showed feasibility and effectiveness of the proposed method

关 键 词：模糊CMAC 混沌优化 遗传算法 机器人 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]