摘要
利用递归神经网络对单框架控制力矩陀螺(SGCMG)系统操纵律进行动态求解,设计了一种基于递归神经网络的SGCMG系统操纵律。通过选择适当参数可以使该网络渐近收敛到稳定状态,从而使操纵律具有较小的操纵误差。该操纵律不用计算Jacobi矩阵的伪逆,因此避免了Jacobi矩阵求逆所带来的一系列问题。对某SGC-MG系统的仿真结果表明,上述操纵律是可行的。
The SGCMG (Single Gimbal Control Moment Gyros) steering law using recurrent neural network is developed. A specific recurrent network was constructed to solve the problem dynamically. With properly chosen parameters, the network could converge to stable states asymptotically, hence the steering error was limited to very small. Without using the pseudoinverse of the Jacobi matrix, the presented method avoids the related calculation problems. Finally, simulation results of a 4 - SGCMG system indicate that the proposed algorithm works well.
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2008年第6期1908-1911,共4页
Journal of Astronautics
基金
国家自然科学基金(10772011)
航天科技创新基金
关键词
控制力矩陀螺
递归神经网络
航天器
Control moment gyros
Recurrent neural network
Spacecraft