Distributed decentralized control for very large-scale systems with application to LEO satellite mega-constellations

Leonardo Pedroso

Key information

Authors:

Leonardo Pedroso (Leonardo Pedroso)

Supervisors:

Pedro Tiago Martins Batista (Pedro Tiago Martins Batista)

Published in

11/17/2022

Abstract

The advantages and tremendous potential of very large-scale complex networks of interconnected systems are indisputable in a myriad of engineering fields, not only as business opportunities but also as a natural change towards efficiency, reliability, and scalability. In particular, mega-constellations of satellites promise to revolutionize the future of communications and Earth observation and monitoring. Although efforts towards the deployment of these solutions are underway, decades-old tried and tested individual ground-based tracking telemetry and command technologies condemn these ventures to practical unfeasibility and economic unviability. The goal of this thesis follows from the self-evident void in the state-of-the-art, aiming to bring these endeavors to fruition. First, the distributed and decentralized control problem is formulated in a receding horizon control framework alongside the severe large-scale feasibility constraints. Second, a convex relaxation procedure is proposed to approximate the optimal solution of the regulator synthesis problem in a decentralized setting, which is validated resorting to a large-scale numeric simulation and experimental results. Moreover, a tracking solution is put forward. Third, a novel distributed and decentralized networked control solution is developed for the particular case of dynamically decoupled systems. The controller synthesis computations are distributed across the network leveraging the proposed convex relaxation, an approximation, and a scheduling procedure, to comply with the feasibility constraints on a very large-scale. Fourth, the potential of the proposed solution is successfully illustrated for the cooperative on-board orbit control of the Starlink mega-constellation. The shape-keeping task is formulated in a novel framework with emphasis on efficiency and fuel saving.