论文部分内容阅读
采用深度学习方法和自编码机的神经网络建模和优化方法,求解空天飞行器的控制分配问题。将空天飞行器控制分配问题的期望控制量看作自编码网络的输入,将舵面实际产生的控制力矩看作自编码网络的输出,通过构建一种特殊形式的深度神经网络,建立自编码机和控制分配问题的等价模型,在不需要用优化算法计算训练样本的前提下,实现了非线性控制分配。提出了一种全新的智能控制分配方法,与早期的基于神经网络的控制分配方法有本质不同。新方法能够很好地处理气动数据的非线性特性,具有较强的工程实用性。
The neural network modeling and optimization method of deep learning method and self-encoder is used to solve the control assignment of spacecraft. The controllability of spacecraft control allocation is considered as the input of self-encoding network. The control moment actually generated on the control surface is regarded as the output of self-encoding network. By constructing a special form of deep neural network, a self- And the equivalent model of the control distribution problem, the nonlinear control distribution is realized without the need of using the optimization algorithm to calculate the training samples. A new intelligent control allocation method is proposed, which is different from the earlier control distribution method based on neural network. The new method can well deal with the nonlinear characteristics of aerodynamic data and has strong engineering practicability.