Date of Award

12-2026

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Aerospace, Physics, and Space Sciences

First Advisor

Madhur Tiwari

Second Advisor

Seong Hyeon Hong

Third Advisor

Vivek Muralidharan

Fourth Advisor

Donald Platt

Abstract

Cislunar missions have gained significant attention in recent decades, motivating the need for efficient modeling and reliable control. In this work a Koopman operator based framework is developed for approximating the error dynamics around a reference Near Rectilinear Halo Orbit (NRHO) in the Earth-Moon Circular Restricted Three-Body Problem (CR3BP). A decoder free neural network is used to learn a lifted linear representation of the nonlinear CR3BP dynamics and a residual based approach is used to identify the corresponding control input matrix. The model is then implemented in a receding-horizon target point controller and compared with uncontrolled propagation and a State Transition Matrix (STM) baseline. For the chosen control settings the Koopman controller has lower tracking errors than the STM controller in most cases. These results show that the learned Koopman model can provide useful short-horizon prediction and control around the selected NRHO.

Download

Share

COinS