CAJ | 학술논문

将小脑模型关节控制器（CMAC）神经网络应用于动载体光电稳定跟踪控制系统设计，分别构建CMAC学习算法网络和CM AC控制网络，泛化参数取4，采用δ学习算法调整网络权值，为评估所构建的CM AC网络对目标系统的逼近能力，选定一个非线性系统作为对象，以连续方波为输入信号进行仿真。仿真数据显示，输入信号发生跳变经0.15 s后输出信号的稳态误差为0。选用直流力矩电机和分辨率为767×10－6 rad的光电编码器构建动载体三轴姿态稳定控制实验装置。结果表明，构建的以CM AC神经网络为核心的控制器在此实验装置上实现的姿态稳定误差为870×10－6 rad。
장소뇌모형관절공제기（CMAC）신경망락응용우동재체광전은정근종공제계통설계，분별구건CMAC학습산법망락화CM AC공제망락，범화삼수취4，채용δ학습산법조정망락권치，위평고소구건적CM AC망락대목표계통적핍근능력，선정일개비선성계통작위대상，이련속방파위수입신호진행방진。방진수거현시，수입신호발생도변경0.15 s후수출신호적은태오차위0。선용직류력구전궤화분변솔위767×10－6 rad적광전편마기구건동재체삼축자태은정공제실험장치。결과표명，구건적이CM AC신경망락위핵심적공제기재차실험장치상실현적자태은정오차위870×10－6 rad。
The cerebellar model articulation controller (CMAC) neural network is applied to the design for photoelectric stable tracking control system of motorial carrier ,respectively constructing “CMAC network learning algorithm” and “CMAC control network” .The generalization parameter is 4 ,and the learning algorithm δis used to adjust to the network weights .In order to assess the approximation ability of the CMAC network to built a target system ,a nonlinear system is selected as the object .The input signal is continuous square wave and simulated .The simulation data shows that the input signal has changed after 0.15 seconds and the steady‐state error of the output signal is 0 .Using DC torque motor and resolution of 767 × 10-6 rad of photoelectric encoder ,it builds a three‐axis attitude stability control experimental device of motorial carrier .The result shows that controller based on CMAC neural network is constructed which can realize attitude stabilization error as 870 × 10-6 rad in this experiment device .