自组织小脑神经网络算法研究

Method of Self-organization CMAC Neural Network

针对神经网络逼近非线性函数问题,提出一种在线的自组织小脑神经网络,其不需要预先确定存储空间的大小。该网络可以根据输入数据自适应地改变神经网络节点数和相对应的权值,具有良好的智能性;针对挠性卫星姿态控制,将自组织小脑神经网络用于不确定项的补偿中,提高系统的鲁棒性。仿真结果表明,该方法神经网络性能良好,实现了卫星姿态控制。

To solve the problem of neural network approaching non-liner function, a new on-line self-organization CMAC neural network is proposed,of which the memory need not to be predefined. The self-organization CMAC neural network can update structure and weights automatically. The paper use sliding mode control to solve the problem of attitude maneuvering of flexible satellite, using self-organization CMAC neural network to compensate the uncertainty. Practical example shows that the proposed approach applied to control the satellite and the network can provide zood performance.