Distributed decentralized receding horizon control for very large-scale networks with application to satellite mega-constellations

TL;DR

This paper proposes a scalable decentralized receding horizon control method for large networks, specifically applied to satellite mega-constellations, addressing communication delays and computational constraints.

Contribution

It introduces a distributed control approach that handles delays, uses local communication, and scales independently of network size, suitable for very large systems.

Findings

Effective control of Starlink mega-constellation demonstrated

Method handles communication delays and local communication

Computational requirements do not grow with network size

Abstract

The implementation feasibility of control algorithms over very large-scale networks calls for hard constraints regarding communication, computational, and memory requirements. In this paper, the decentralized receding horizon control problem for very large-scale networks of dynamically decoupled systems with a common, possibly time-varying, control objective is addressed. Each system is assumed to be modeled by linear time-varying dynamics, which can be leveraged to approximate nonlinear systems about successive points of operation. A distributed and decentralized receding horizon control solution is put forward, which: i) takes communication delays into account; ii) allows local communication exclusively; and iii) whose computational and memory requirements in each computational unit do not scale with the dimension of the network. The scalability of the proposed solution enables emerging very large-scale applications of swarm robotics and networked control. This approach is applied to the orbit control problem of low Earth orbit mega-constellations, featuring high-fidelity numerical simulations for the Starlink mega-constellation.