SCORE оптимизирует расположение наземной станции LEO на 15% больше Downlink

Up to 15% greater total downlink throughput — that’s what you get when you stop picking ground station sites from a list and instead let an optimizer place them anywhere on Earth’s surface.

That’s the headline result from a new paper introducing SCORE (Sequential Cyclic Optimization via Refinement & Evaluation), a free-placement method for ground station network design targeting LEO satellite constellations. Current practice restricts sites to existing infrastructure — a small set of predefined locations from providers like Kongsberg Satellite Services or the World Teleport Association. SCORE operates over a continuous spatial domain, which is a harder optimization problem (high-dimensional, non-convex, littered with local minima) but yields a materially better network.

Free Placement Beats Fixed-Site by 15%

The authors benchmarked SCORE against two commercial Earth observation constellations — Capella Space and ICEYE — plus one synthetic Walker-Star constellation. Against fixed-site methods (integer programming over existing locations), unconstrained SCORE achieved up to 15% greater total downlink throughput. Even when they constrained placement to within proximity of existing fiber and power infrastructure, the method retained over 92% of that gain. That means you don’t need to build in the middle of nowhere to capture most of the benefit.

SCORE’s Two-Stage Optimization Beats Global Optimizers

SCORE combines sequential coordinate selection with cyclic refinement — a two-stage approach that tackles the high-dimensionality and non-convexity that choke one-shot global optimizers like differential evolution (DE). The empirical results are stark: SCORE requires up to 5x fewer function evaluations to converge relative to DE while simultaneously improving downlink throughput by up to 13%. That’s not a marginal efficiency gain; it’s the difference between a solver that finishes overnight and one that runs for days.

Infrastructure Constraints Don’t Kill the Gains

One practical question: how much of the throughput advantage evaporates when you have to build near existing roads, power, and fiber? The paper quantifies this trade-off explicitly. Infrastructure-constrained SCORE still delivers 92%+ of the unconstrained throughput improvement over fixed-site baselines. That shifts the ground network design conversation from “should we expand existing stations or build new ones?” to a data-driven cost-benefit analysis where deploying new sites near existing infrastructure is now clearly justified.

SCORE establishes an empirical performance benchmark for flexible placement that competing methods will have to match. For operators like Capella and ICEYE running growing constellations, the next step is integrating this optimizer into their actual deployment planning pipelines.

Source: Free-Placement Optimization of Ground Station Locations for Low-Earth Orbit Satellites
Domain: arxiv.org