Dissertation/Thesis Abstract

Variable Memory Recurrent Neural Networks for Nano Sat Launch Vehicle attitude control
by Sclafani, Rodolfo J., M.S., California State University, Long Beach, 2014, 79; 1527749
Abstract (Summary)

Launch Vehicles are governed by a complex set of nonlinear and highly coupled differential equations. In general, these equations are linearized about an equilibrium point and a linear controller is designed based upon these linearized dynamics. The linear controller is then applied to the actual system which is, of course, the original nonlinear system. This, in turn, leads to tracking errors and poor performance when the vehicle experiences significant deviation from the equilibrium condition. Also, the linear controller has difficulty handling unknown variations in the system parameters which also give rise to poor performance when applied to the actual vehicle.

A new type of neural network known as a Variable Memory Recurrent Neural Network (VMRNN) has been designed that, when used in conjunction with a linear PID controller, offers improved transient response characteristics when encountering uncertainty in the dynamic model and external disturbances.

Indexing (document details)
Advisor: Shankar, Praveen
School: California State University, Long Beach
Department: Mechanical and Aerospace Engineering
School Location: United States -- California
Source: MAI 52/06M(E), Masters Abstracts International
Subjects: Aerospace engineering
Keywords: Attitude control, Launch vehicle, Neural network, Recurrent neural networks
Publication Number: 1527749
ISBN: 978-1-303-92605-1
Copyright © 2021 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy