机器人柔顺装配夹具的双小脑模型神经网络控制系统  

Dual CMAC Neural Network Control System of Compliant Fixture for Robotic Assembly

作　　者：颜波[1] 张铁[1] 谢存禧[1] 

机构地区：[1]华南理工大学机电工程学院,广州510640

出　　处：《高技术通讯》2002年第12期53-56,共4页Chinese High Technology Letters

基　　金：中国科学院机器人学开放研究实验室基金 (RL19990 4);广东省自然科学基金 ( 0 0 0 478)资助项目。

摘　　要：介绍了一种新型的机器人柔顺装配夹具及其装配系统。设计了由装配力信号为输入的小脑模型神经网络CMAC1和以浮动平台位移信号为输入的小脑模型神经网络CMAC2并行叠加构成的双CMAC神经网络控制器 ,前者作为前馈控制 ,后者当夹具刚启动时为前馈控制 ,装配开始后转为反馈控制。双CMAC神经网络控制器具有存储容量少、系统较稳定、能减少系统误差影响、装配力小等优点。A new type of compliant fixture for robotic assembly and its assembly system are introduced. Dual CMAC neural network controller is designed, which is composed of cerebellar model articulation controller CMAC1 that receives the assembly force signals and cerebellar model articulation controller CMAC2 that receives floating platform displacement signals. The former is feed-forward control, while the latter is feed-forward control when the fixture just start and then converts to feedback control after assembly begins. Dual CMAC neural network controller have advantages such as less memory volume, higher stability, less influences from system errors and less assembly force.

关 键 词：小脑模型神经网络 控制系统 机器人 柔顺装配夹具 小脑模型关联控制器 CMAC 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]