一种基于网点的小脑模型及其在非线性动态系统建模中的应用

A CMAC WITH A NET POINT STRUCTURE AND ITS APPLICATION TO NONLINEAR DYNAMIC SYSTEM MODELING

小脑模型（ＣｅｒｅｂｅｌａｒＭｏｄｅｌＡｒｔｉｃｕｌａｔｉｏｎＣｏｎｔｒｏｌｅｒ，ＣＭＡＣ）常用于学习控制，可看作一种基函数网络。本文提出了网点的概念和以网点为中心的超闭球量化方法，给出了快速学习算法，并分析了其收敛性，通过非线性动态系统连续搅拌釜反应器的建模仿真研究，结果表明该ＣＭＡＣ具有很强的学习记忆功能，可用于非线性动态建模。

The cerebellar model articulation controller (CMAC) is often used in learning control system, it can be viewed as a basis function network. A conception of net points and a netpointbased CMAC with a novel hyperball quantization method and its high speed learning algorithm are presented, and the convergence of learning is analyzed. Simulations for the CMAC used in nonlinear dynamic system modeling are performed to demonstrate its powerful associative memory performance and the accuracy in modeling.