摘要
小脑模型神经网络(CMAC)用作前馈控制器的结构对于含有丰富状态的轨迹跟踪问题,如机器人轨迹跟踪问题有较好的控制效果,但它对于参考值为方波或阶跃信号的跟踪问题却不太适合.为此,提出CMAC网络反馈控制器的方案,该方案克服了CMAC局部联想特性和前馈控制结构所固有的缺陷,对较恒定的设定值跟踪和抑制噪声都有较好的效果.最后,仿真实例说明了所提方案的有效性.
Using the Cerebellar Model Articulation Controller(CMAC) as the feed forward controller, Miller successfully implemented the robot trajectory learning control. Because of the built-in shortcomings of the local-generalization of CMAC and the feedforward control structure, the scheme is not suitable for constant regulator and noise disturbance. The feedback structure of CMAC is proposed. The simulation result shows the effectiveness of the strategy.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1996年第4期114-118,共5页
Journal of Shanghai Jiaotong University
关键词
神经网络
反馈控制
学习控制系统
s: neural network
feedforward
feedback
cerebellar model articulation controller