• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>AIAA guidance, navigation, and control conference;AIAA SciTech forum >Variable Memory Recurrent Neural Networks For Launch Vehicle Attitude Control
【24h】

Variable Memory Recurrent Neural Networks For Launch Vehicle Attitude Control

机译：用于运载火箭姿态控制的可变记忆递归神经网络

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Launch vehicle attitude control is a complex nonlinear problem that is compounded by several internal and external factors such as modeling uncertainties, elastic dynamics, fuel sloshing and unstable aerodynamics. Adaptive control strategies that utilize a neural network augmented to a linear controller have the capability to minimize the effect of these uncertainties and disturbances. In this paper, the architecture and learning mechanism of a new type of recurrent neural network known as the Variable Memory Recurrent Neural Network (VMRNN) is presented. This network, when used in conjunction with a proportional integral derivative (PID) controller for the attitude control of a Nanosat Launch Vehicle, offers improved transient response characteristics when encountering uncertainty in the dynamic model and external disturbances. Further, a comparison of the VMRNN augmented PID controller with a traditional Feedforward Neural Network (FFNN) and Recurrent Neural Network (RNN) augmented PID controller demonstrates its improved learning capability.
机译：运载火箭的姿态控制是一个复杂的非线性问题，其中包括一些内部和外部因素，例如模型不确定性，弹性动力学，燃油晃荡和不稳定的空气动力学。利用扩展为线性控制器的神经网络的自适应控制策略具有将这些不确定性和干扰的影响降至最低的能力。本文提出了一种新型的递归神经网络，称为可变记忆递归神经网络（VMRNN）的体系结构和学习机制。当与比例积分微分（PID）控制器一起用于Nanosat运载火箭的姿态控制时，该网络在遇到动力学模型的不确定性和外部干扰时可提供改进的瞬态响应特性。此外，VMRNN增强型PID控制器与传统前馈神经网络（FFNN）和递归神经网络（RNN）增强型PID控制器的比较证明了其改进的学习能力。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号