【摘要】 基于牛顿前向插值公式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———NFI CMAC,详细讨论了其插值算法、训练规则及寻址机制.与传统CMAC相比,NFI CMAC具有学习精度高、学习速度快及占用内存单元少等优点.基于NFI CMAC设计了一种高性能的机器人轨迹跟踪分布式智能控制方案,仿真研究表明了该方案的可行性与有效性.更多还原

【Abstract】 This paper proposes a novel high-order CMAC_type neural network via the Newton’s forward interpolation (NFI-CMAC), which is capable of implementing error-free approximations to multi-variable polynomial functions of arbitrary order, including interpolation algorithm and training algorithm. Compared with the conventional CMAC-type AMS, the proposed one has advantages such as high-precision of learning, much smaller memory requirement without the data-collision problem. Based on the proposed neural network, the authors design a type of simple and truly general robotic manipulator intelligent controller; the simulation results verify that the control strategy is effective.更多还原