基于新型CMAC神经网络的复杂系统控制

首先基于离散泰勒级数设计了一种可实现对任意阶多维函数无差逼近的新型CMAC神经网络,给出了其插值算法及训练算法。该系统相对于传统的CMAC神经网络具有学习精度高、收敛速度快、所需内存单元少等优点。并基于新型CMAC神经网络设计了一种复杂系统的高性能实时控制策略。该方案不依赖于系统的数学模型,具有较强的鲁棒性和通用性,仿真结果表明了该方案的可行性与有效性。

Complex system control strategy based on a novel CMAC-type neural network

On the basis of discrete Taylor series,this paper proposes a novel CMAC neural networkor (DTS CMAC), including an interpolation algorithm and a training algorithm, which is capable of implementing error free approximations to multi variable polynomial functions of arbitrary order. The network offers several advantages such as high learning precision and much smaller memory requirement without data collision problem compared to the conventional CMAC-type neural network.Based on DTS CMAC,a complex on line system control strategy with high performance,independent on system mathematic model,robust,and universal is designed.Simulation results verify that this proposed control strategy is both feasible and efficient.