02 · RESEARCH TOPIC
Mega-Constellations
Thousands of satellites linked by lasers form a network that reshapes itself every second. I work on routing through that mesh, on pointing and holding the lasers, and on who these networks actually serve.
Constellations such as Starlink now fly thousands of satellites a few hundred kilometres up. Each one crosses the sky in minutes, and many are linked to their neighbours by lasers, so the network behind the service changes shape all the time.
With students and colleagues, I work on routing data through that moving mesh, on getting laser terminals to find each other and hold the link, and on how operators can run and design fleets of this size, including with the help of AI. With UC Berkeley, I also look at who these networks reach, and at what price.
Across the mesh
A route between two cities through a model of Starlink's first shell, with the size, height and tilt used in our Starlink routing study: 1,584 satellites in 72 planes of 22, 550 km up and inclined 53°, each with laser links to the satellites ahead of and behind it and to one in each neighbouring plane. The route climbs to a satellite over the first city, crosses the mesh and comes down over the second. Its delay is set against a fibre laid straight along the ground between them; in glass, light travels at about two thirds of its speed in vacuum. Distributed On-Demand Routing, ASMS/SPSC 2022, §II-C and §VI-A
- Laser hops–
- Path length–
-
One-way delay, light only
Through the mesh –
Straight fibre –
Model: one shell of circular orbits; four laser links per satellite (three on the first and last planes, which are not linked to each other); each city reaches any satellite at least 25° above its horizon, and the route is the shortest by length. Light only: no queues, no ground network and no processing on board, which the routing study compares to about 1,000 km of light travel, 3.3 ms, per hop (§IV-B). Real cables run longer than the great circle.
Results, with sources
- Satellites to reach every under-served address. About 4.7 million US addresses have no provider offering broadband at 100 Mbps down and 20 Mbps up. Serving all of them without overselling capacity beyond the 20-to-1 ratio that US regulators allow fixed wireless providers would take Starlink over 40,000 satellites, more than five times the roughly 8,000 it had at the time of the study. The authors present this as a lower bound. Anyone, Anywhere, HotNets 2025, §3, Table 2
- Who can pay for it. At $120 a month, Starlink's residential plan would cost more than 2% of household income, a common affordability threshold, at about 3.5 million of those 4.7 million addresses, and at nearly 3 million even with the federal Lifeline discount. Comparable cable plans at $40 to $50 a month are affordable at more than 99.99% of them. The study uses each county's median income. Anyone, Anywhere, HotNets 2025, §4, Fig. 4
- Getting lasers to lock on. Before a laser link between two satellites can carry data, both must turn to face each other and find the beam. In simulations of two 12 kg satellites, that left only 72% of a contact for data on the harder of two passes. Turning early and exchanging satellite-navigation positions over a radio link raised the share to over 99%, and cut the memory needed to hold waiting data at 100 Mbps from 4 GB to 75 MB. RF-Assisted Compensation, IEEE OJ-COMS 2026, §IV-B, Tables 8 and 9
- Finding a path through the mesh. On a model of Starlink's first shell of 1,584 satellites, each linked to its four neighbours, a routing method that first works out the fewest satellite-to-satellite hops found paths on average 158 times faster than the textbook shortest-path algorithm, and the routes it finds are nearly as short. The speed-up is in the time a computer takes to work out a route. Distributed On-Demand Routing, ASMS/SPSC 2022, §VI-B and §VI-C, Figs. 5 and 6
The last two are simulation results, under the assumptions set out in each paper. The first two come from a model built on public data from the US regulator (FCC).
Tools and projects
-

D3-CONNECT
Direct-to-device satellite connectivity for everyone, with Sylvia Ratnasamy's NetSys Lab at UC Berkeley.
Principal investigator
-

ConOpsCon
Constellation operations with continuous communication, with Saarland University as prime contractor and VisionSpace Technologies.
Coordinator
-

DONUTS
Design of multiscale, multitechnology networks: a seamless continuum of space, air and ground networks for 5G and 6G.
Contributor
-

AI4LFNM, ML4OISL and HANDING-OVER
AI for managing large satellite fleets, machine learning for optical links between satellites, and handover and routing in very low Earth orbit, with GMV, Thales Alenia Space, Eutelsat, Planet Labs and Politecnico di Torino.
Lead researcher (AI4LFNM) · contributor
Selected papers
-
Anyone, Anywhere, not Everyone, Everywhere: Starlink Doesn't End the Digital Divide
ACM Workshop on Hot Topics in Networks (HotNets), 2025
Abstract
Low Earth Orbit (LEO) satellite constellations, such as Starlink, are increasingly promoted as a solution to the digital divide in rural and underserved communities. In this paper, we take a closer look at the limits of this approach. Using the insight that capacity limitations of LEO-based access networks are driven by peak demand density, we introduce a simple analytical model that brings together real-world demand data with the physical and regulatory limits of LEO satellite networks. Applying our model to broadband demand across the United States, we find that serving the current Starlink constellation size is likely insufficient for covering all un- and underserved locations in the US and we find diminishing returns that disincentivize scaling the constellation to serve the long-tail of these un(der)served locations. We also identify that Starlink's current pricing is likely unaffordable for the majority of these locations, even with existing government subsidies. We argue that LEO constellations, while technologically impressive, are just another piece of the solution, rather than a panacea. New, innovative approaches are still required to end the digital divide.
-
On the Use of Mega Constellation Services in Space: Integrating LEO Platforms Into 6G Non-Terrestrial Networks
IEEE Journal on Selected Areas in Communications, 2024
Abstract
This paper presents a framework for integrating Low-Earth Orbit (LEO) platforms with Non-Terrestrial Networks (NTNs) in the emerging 6G communication landscape. Our work applies the Mega-Constellation Services in Space (MCSS) paradigm, leveraging LEO mega-constellations’ expansive coverage and capacity, designed initially for terrestrial devices, to serve platforms in lower LEO orbits. Results show that this approach overcomes the limitation of sporadic and time-bound satellite communication links, a challenge not fully resolved by available Ground Station Networks and Data Relay Systems. We contribute three key elements: (i) a detailed MCSS evaluation framework employing Monte Carlo simulations to assess space user links and distributions; (ii) a novel Space User Terminal (SUT) design optimized for MCSS, using different configurations and 5G New Radio Adaptive Coding and Modulation; (iii) extensive results demonstrating MCSS’s substantial improvement over existing Ground Station Networks and Data Relay Systems, motivating its role in the upcoming 6G NTNs. The space terminal, incorporating a multi-system, multi-orbit, and software-defined architecture, can handle Terabit-scale daily data volumes and minute-scale latencies. It offers a compact, power-efficient solution for properly integrating LEO platforms as space internet nodes.
-
RF-Assisted Uncertainty Cone Reduction in Free-Space Optical Inter-Satellite Links
IEEE Open Journal of the Communications Society, 2024
Abstract
One of the space communications industry’s current focuses is developing high-throughput communication terminals for satellite-to-satellite communication links. Optical inter-satellite links provide high data rates, long-range, and robustness against interferences, and they do not require frequency licensing as radiofrequency communication systems. Nevertheless, as the uncertain location of the receiver is comprised in an area larger than that illuminated by the transmitting laser, the pointing accuracy is a critical element in the success of the optical link establishment and maintenance, requiring a pointing, acquisition, and tracking mechanism. The acquisition process is the most time-consuming of the pointing processes, limiting the time available to send data, especially in highly dynamic networks. This paper focuses on reducing the acquisition time by reducing the initial satellite position uncertainty. To this end, a hybrid system that combines RF and optical technologies in a single communication module is proposed. Whereas the control plane is managed via the RF link to exchange more precise global navigation data, the optical link corresponds to the data plane in which payload data is exchanged. The pointing between two satellites is simulated to analyze its behavior, considering the error of cumulative orbital propagation data and global navigation satellite system data. This work also analyzes the cumulative error produced by the propagation of the TLEs over time. Finally, the results show how a system that relies on the exchange of global navigation satellite system positioning data achieves up to 99.45% better-pointing accuracy than a system that bases positioning data on TLE propagation.
-
Distributed On-Demand Routing for LEO Mega-Constellations: A Starlink Case Study
Advanced Satellite Multimedia Systems Conference and Signal Processing for Space Communications Workshop (ASMS/SPSC), 2022
Abstract
The design and launch of large-scale satellite networks create an imminent demand for efficient and delay-minimising routing methods. With the rising number of satellites in such constellations, pre-computing all shortest routes between all satellites and for all times becomes more and more infeasible due to space and time limitations. Even though distributed on-demand routing methods were developed for specific LEO satellite network configurations, they are not suited for increasingly popular mega-constellations based on Walker Delta formations.The contributions of this paper are twofold. First, we introduce a formal model that mathematically captures the time-evolving locations of satellites in a Walker Delta constellation and use it to establish a formula to compute the minimum number of ISL hops between two given satellites. In the second part, we present an on-demand hop-count-based routing algorithm that approximates the optimal path while achieving superior performance compared to classical shortest-path algorithms like Dijkstra.
-
Managing Fleets of LEO Satellites: Nonlinear, Optimal, Efficient, Scalable, Usable, and Robust
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2020
Abstract
Size and weight limitations of low-earth orbit (LEO) small satellites make their operation rest on a fine balance between solar power infeed and power demands of communication technologies on board, buffered by on-board battery storage. As a result, the problem of planning battery-powered payload utilization together with intersatellite communication is extremely intricate. Nevertheless, there is a growing trend toward constellations and megaconstellations that are to be managed using sophisticated software support. Earlier work has leveraged cost-optimal reachability in priced timed automata for deriving near-optimal finite-horizon schedules to operate a single LEO satellite in orbit. This article harvests that work and improves it in several dimensions, all needed for true in-orbit applicability: 1) the battery representation is no longer bound to be linear, but can be kinetic, which means that the optimization problem includes nonlinearities; 2) the management is perpetuated by a receding horizon scheduling strategy; 3) the model is continuously improved with the latest telemetry received from orbit; 4) a tandem of satellites equipped with state-of-the-art intersatellite link transponders is considered; 5) the core optimization problem is now solved using dynamic programming with antichain-based pruning, which is proven to be optimal and despite all the additional features outperforms the earlier approach by orders of magnitude; 6) the entire approach is grounded in the concrete requirements of the GOM X–4 LEO mission; 7) care is taken to make the approach usable by the space engineers, and robust against failures of parts of the toolchain; and 8) an extensive test campaign validates accuracy, efficiency, scalability, and robustness with respect to the operational requirements and constraints of LEO constellations.
People and partners
- Gregory F. Stock, PhD student at Saarland University, co-supervised with Holger Hermanns: routing in mega-constellations and operating satellite fleets.
- Elena F. Niño, PhD student at i2CAT and UPC, co-supervised with Joan A. Ruiz-de-Azua: pointing and acquisition for laser links between satellites.
- Arnau Singla, PhD student at i2CAT and UPC, co-supervised with Joan A. Ruiz-de-Azua: scheduling constellations together with mobile networks on the ground.
- Santiago Henn, PhD at UNC (2026): coverage computation, segmented satellite architectures and TCP over Starlink.
- Gabriel Maiolini Capez, PhD at Politecnico di Torino (2026), co-supervised with Roberto Garello: routing in low and very low orbits, and satellites that use mega-constellations as a service.