摘要
基于离散泰勒级数提出一种对多维函数可实现任意阶逼近的新型CMAC神经网络———DTS CMAC ,详细讨论了该系统的插值算法、训练规则 ,与传统CMAC相比 ,DTS CMAC具有学习精度高、学习速度快及占用存储单元少等优点。基于DTS CMAC设计了一种高性能的机械手轨迹跟踪分布式智能控制方案 ,并以肘关节为例 ,设计了机械手关节转矩控制器 。
A kind of new typed CMAC neural network——DTS-CMAC, which could achieve arbitrary ordered approach, has been put forward based on discrete Taylor series, and the interpolation algorithm and training rule of that system has been discussed. Comparing with traditional CMAC, the DTS-CMAC possesses advantages of high learning precision, quick learning speed and occupying less memory unit. A kind of distributed intellectual controlling scheme for tracing the manipulator path was designed based on DTS-CMAC. And taking the elbow joint as an example, the torque controller of manipulator′s joint was designed, the study of simulation indicated the feasibility and effectiveness of that scheme.
出处
《机械设计》
CSCD
北大核心
2004年第12期30-31,43,共3页
Journal of Machine Design
基金
教育部科学技术重点资助项目 (0 2 0 1 2 )
关键词
离散泰勒级数
神经网络
机械手
智能控制
discrete Taylor series
neural network
manipulator
intellectual control
