Multi-Objective Rendezvous Operation Design Using Neural ODEs with Reservoir Computing Architecture for Trajectory Control Laws

Trajectory control laws based on neural ordinary differential equations (ODEs) were proposed by the authors in a previous study. This paper extends the trajectory control laws by integrating a reservoir computing architecture. The addition can significantly reduce the number of design parameters to...

Full description

Saved in:
Bibliographic Details
Published in2024 SICE Festival with Annual Conference (SICE FES) pp. 63 - 68
Main Authors Ueda, Satoshi, Ogawa, Hideaki
Format Conference Proceeding
LanguageEnglish
Published The Society of Instrument and Control Engineers - SICE 27.08.2024
Subjects
Computer architecture
Design optimization
Low earth orbit satellites
multi-objective design optimization
neural ODE
Optimization
Orbits
Ordinary differential equations
Reservoir computing
System analysis and design
Training
Trajectory
trajectory control
Online AccessGet full text

Cover

More Information
Summary:Trajectory control laws based on neural ordinary differential equations (ODEs) were proposed by the authors in a previous study. This paper extends the trajectory control laws by integrating a reservoir computing architecture. The addition can significantly reduce the number of design parameters to be optimized and thus the computational cost of training while maintaining the structure of the control laws. This enhancement allows general-purpose nonlinear optimization algorithms to be used, extending the application of neural ODEs to operation design optimization beyond the design of control laws. This study verifies the effectiveness of the proposed framework by constructing a control law to achieve the target conditions for a rendezvous operation in a low Earth orbit. Multi-objective design taking advantage of reservoir computing is demonstrated by assessing the Pareto optimal front with respect to orbital transfer time and control accuracy.