Starlink satellites moving in front of the Pleides and Venus. Credit: T. Hansen/IAU OAE/Creative Commons Attribution

Satellite mega-constellations are quickly becoming the backbone of a number of industries. Cellular communication, GPS, weather monitoring and more are now, at least in part, reliant on the networks of thousands of satellites cruising by in low Earth orbit. But, as these constellations grow into the tens of thousands of individual members, the strain they are putting on the communications and controls systems of their ground stations is becoming untenable.

A new paper from Yuhe Mao of the Nanjing University of Aeronautics and Astronautics and their co-authors published in Space: Science & Technology, hopes to alleviate some of that pressure by offloading much of the control scheme and network decision-making logic to satellites themselves.

In traditional satellite mega-constellations, the ground control units are responsible for communicating with each individual satellite. This leads to bottlenecks in both processing power and communications channel bandwidth as the number of satellites each ground station is responsible for grows. Those bottlenecks are then reflected in increased latency times, which can be a death knell for constellations that get too large, as latency is one of the primary metrics by which end users judge a communications network.

Designing a system that off-loads the control and networking decisions from the ground station sounds relatively simple, but in practice it is much harder to implement. The authors utilized a technique called Software Defined Networking (SDN) to move the decision for the control layer up to a series of satellites they called “Center Nodes.” Each of these Center Nodes would be responsible for communicating both with the ground stations set up to support the constellation, but also with all the “Member Nodes” (i.e., other physically identical satellites) in their general area.

Center Nodes are selected early in the constellation’s lifetime, and are physically the same as all the rest of the satellites in the constellation. However, they are told to operate in a different mode by the SDN algorithm due to their opportune placement in orbit. These are the only satellites allowed to communicate with the ground, in an effort to limit the data traffic and control complexity flowing through the ground links to the constellation.

Member satellites, on the other hand, are responsible for finding and connecting to whatever Center Node is the “best fit” for them. Importantly, this doesn’t always mean the closest. The authors lay out a “construction algorithm” which each member satellite is responsible for calculating for themselves, that takes into account factors like how long it will take until a manager satellite drifts out of range. Being “out of range” in this context means moving farther away than half of the maximum communications distance possible between two satellites.

That calculation of the amount of time before a satellite will have to switch who is managing it, the “Detachment Time,” is arguably the most important part of the paper. Typically, satellites would have to solve a multivariable calculus problem using a system called propagation to determine how long it would take to get out of range of a manager and into the range of another. Using this technique, it would take factors like the current manager’s position and velocity values in the three dimensions and calculate where they would end up for the next using an integration algorithm.

Alternatively, the new algorithm uses a “prediction” algorithm, which only requires the calculation of what the paper calls the “geocentric angle” between the two satellites. Since their orbits are well-defined, it’s a relatively simple algebraic equation to solve for the relative speeds of the two satellites compared to one another, which can then be used to determine the amount of time until the distance between them breaches the maximum distance threshold.

Importantly, all of this math is able to be done for multiple potential “managers” at one given time, even with the constrained computational power onboard the satellites themselves. And, also importantly, the manager selection algorithm penalizes unnecessary switching, so the number of “switches” of a satellite to a different manager is minimized.

To prove all this out, the authors built a simulation of 1,248 satellites based on the orbital characteristics of some early Starlink versions. The networking algorithm split the constellation into 81 separate “management domains,” each handled by a single manager satellite. They then simulated a month of the constellation’s operation, and noted that, on average, only about six satellites changed manager links per hour. But, perhaps more impressively, the latency averaged between 4.7 and 7.8 ms, compared to a latency of 18.4 ms for a simulation that was run without this maintenance algorithm.

Cutting latency to a satellite mega-constellation by more than 50% is an impressive feat. But, to be clear, this hasn’t been implemented in any actual constellation hardware yet, and it still needs some development for edge cases. For example, the algorithm to determine what manager to connect to doesn’t consider how much load is already on that manager, leaving open the possibility of potentially overloading a manager satellite rather than a ground station.

But that’s a simple enough factor to add to the calculation, and the benefits of improved latency seem to far outweigh the risks. It remains to be seen if this algorithm is eventually adopted by Kuiper, Starlink, or any of their ilk. But for now, it simply represented a marked theoretical improvement to an infrastructure component that continues to grow in importance.

More information: Yuhe Mao et al, Dynamic Management Topology Construction, Evolution, and Maintenance of Low Earth Orbit Mega-Constellation, Space: Science & Technology (2025). DOI: 10.34133/space.0248

Citation: How mega-constellations are learning to manage themselves (2025, November 21) retrieved 21 November 2025 from https://phys.org/news/2025-11-mega-constellations.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.