Date of Award

5-2024

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Aerospace, Physics, and Space Sciences

First Advisor

Madhur Tiwari

Second Advisor

Siddhartha Bhattacharyya

Third Advisor

Manasvi Lingam

Fourth Advisor

Ratneshwar Jha

Abstract

The computational and complexity burden of current linearization techniques is one that is a hinderance in the application of real world guidance, navigation and control systems. With the advancements in Deep Neural Networks, large data handling and Koopman Theory, the possibility of global linearizations of nonlinear systems is more prominent. This work demonstrates the capability of a Deep Neural Network learned Koopman operator to transform a nonlinear system into a Linear Time-Invariant system. The method presented is applied to both two purely dynamical systems and one controlled system to emphasize the ability for the technique to be applied in all domains. The Two-Body Problem and Circular Restricted Three-Body Problem are both linearized accurately with a similar learned model, whilst the Pendulum Problem is also accurately linearized with a model that is adapted to include capabilities for control.

Comments

Copyright held by author.

Download

Share

COinS