Optimised LISL Routing Achieves Maximum Throughput for LEO Mega-Constellations

Researchers are tackling the challenge of optimising data flow within Low Earth Orbit (LEO) mega-constellations, which promise to significantly enhance global wireless coverage and resilience. Zhouyou Gu from Singapore University of Technology and Design, Jihong Park from Singapore University of Technology and Design, and Jinho Choi from the University of Adelaide, et al., present a novel approach to jointly manage laser inter-satellite link connections and traffic routing. Their work addresses a critical gap in current schemes, which often fail to account for the limited capabilities of laser communication terminals and uneven global data demands, resulting in inefficient network performance. By formulating the problem using Lagrangian duality, the team decomposes a complex optimisation task into manageable components, demonstrably improving network throughput in simulations using real-world data and offering a pathway to more effective LEO constellation operation?

This innovative use of multipliers allows the system to adapt to changing network conditions and traffic patterns. Experiments show that the proposed method substantially improves network throughput, achieving gains of up to 35%, 145% over existing non-joint approaches. The core innovation lies in the simultaneous optimisation of both physical layer connectivity, establishing the LISLs, and network layer routing, directing the data flow.

By considering the limited steering range and potential vibrations of LCTs, the research ensures that established links are not only high-capacity but also reliably maintainable. This is particularly important given that each satellite typically carries only a limited number of LCTs, making efficient allocation paramount. This breakthrough has significant implications for the future of satellite communication, enabling more resilient and higher-capacity networks for a range of applications. The ability to maximise throughput in LEO mega-constellations is critical for supporting bandwidth-intensive services such as high-definition video streaming, cloud computing, and real-time data analytics. Furthermore, the provable convergence of the subgradient descent optimisation algorithm ensures the stability and reliability of the system, paving the way for practical implementation in future satellite networks and enhancing global connectivity.

Lagrangian duality for LEO constellation optimisation

This innovative approach involved a weighted graph-matching task for LCT connections, weighted shortest-path routing, and a linear program for rate allocation. The core of the study pioneered the use of Lagrange multipliers to reflect congestion weights between satellites, dynamically guiding the matching, routing, and rate allocation processes. Subgradient descent was then implemented to optimize these multipliers, ensuring provable convergence of the algorithm. Experiments employed real-world constellation and terrestrial data to validate the methodology, demonstrating a substantial improvement in network throughput.

Specifically, the team engineered a system where the relaxation of coupling constraints via Lagrangian duality enabled the decomposition of the complex problem into these three distinct, yet interconnected, tasks. This method achieves a significant advancement over existing non-joint approaches by considering both LCT limitations and the global distribution of traffic demand. The weighted graph-matching component strategically assigns LCT connections based on link characteristics, while the weighted shortest-path routing efficiently directs traffic flow across the constellation. The linear program then allocates rates to maximize throughput, all coordinated by the dynamically adjusted Lagrange multipliers. Simulations revealed that the developed methods improve network throughput by 35%, 145% compared to conventional techniques, highlighting the efficacy of the joint optimization strategy. The team’s work demonstrates a practical solution for enhancing the capacity and resilience of future wireless systems reliant on LEO mega-constellations and LISLs.

DuJo optimises LEO throughput via joint routing

Experiments revealed that the developed method, termed DuJo, significantly enhances throughput compared to non-joint approaches. The team measured network throughput using a constellation based on real-world Starlink data from CelesTrak at UTC 2025-07-16 16:00, with varying numbers of satellites. Each satellite was configured with two LCTs, maintaining specific orientations and beam parameters, including an aperture area of 0.01 m2 and a beam wavelength of 1.55μm. Tests confirmed that the beam waist radius was 9.87×10−3m and the Rayleigh range was 1.97×103m, while the pointing jitter was maintained at 10 micro-radians.

The study incorporated realistic user demand, assuming 0.01% of the population within a 200km radius of each satellite required 0.1 Gbps of data. The simulations, conducted on a workstation with an Intel Core Ultra 9 285K processor and 32 GB of memory, show that the dual function value increases and stabilizes over iterations, indicating convergence to an optimal solution. Data shows that network throughput achieved near 500 Gbps with 1000 satellites, surpassing baseline methods like +Grid, Rand, and MRate. Measurements confirm that the network throughput increased with a slower step size decay rate in the subgradient descent optimization, indicating improved exploration of Lagrange multipliers.

Specifically, the team recorded a higher throughput for β = 0.5 compared to β = 0.7 and 0.9. Furthermore, the computing time for the DuJo method remained manageable, with times ranging from 10−3 to 10−1 seconds depending on the number of satellites, demonstrating the scalability of the approach. The breakthrough delivers a pathway to more efficient and resilient LEO mega-constellation networks, enabling high-capacity, long-range connectivity.

Lagrangian duality optimises LEO laser network throughput

Simulations utilising real-world constellation and terrestrial data demonstrate substantial improvements in network throughput, up to and exceeding those achieved by current non-joint optimisation techniques. The iterative optimisation of Lagrange multipliers effectively balances link rates and constellation connectivity, leading to superior LISL matching and flow routing decisions. However, the authors acknowledge that computational time increases with larger constellation sizes, representing a limitation of the current method. Further investigation into the impact of LISL acquisition and tracking processes on overall throughput is also suggested.