让下一篇重要论文，不再错过。

原文摘要

Low-Earth orbit (LEO) mega-constellations are emerging as high-capacity backbones for next-generation Internet. Deployment of laser terminals enables high-bandwidth, low-latency inter-satellite links (ISLs); however, their limited number, slow acquisition, and instability make forming a stable satellite topology difficult. Existing patterns like +Grid and Motif ignore regional traffic, ground station placement, and constellation geometry. Given sparse population distribution on Earth and the isolation of rural areas, traffic patterns are inherently non-uniform, providing an opportunity to orient inter-satellite links (ISLs) according to these traffic patterns. In this paper, we propose Starfield, a novel demand-aware satellite topology design heuristic algorithm supported by mathematical analysis. We first formulate a vector field on the constellation's shell according to traffic flows and define a corresponding Riemannian metric on the spherical manifold of the shell. The metric, combined with the spatial geometry, is used to assign a distance to each potential ISL, which we then aggregate over all demand flows to generate a score for each satellite's link selection. Inspired by +Grid, each satellite selects the link with the minimum Riemannian distance along with its corresponding angular links. To evaluate Starfield, we developed a custom, link-aware, and link-configurable packet-level simulator, comparing it against LAMP, +Grid, Motif, and Random topologies. For the Phase 1 Starlink, simulation results show up to a 30% reduction in hop count and a 15% improvement in stretch factor across multiple traffic distributions. Moreover, static Starfield, an inter-orbital link matching modification of Starfield, achieves a 15% improvement in stretch factor under realistic traffic patterns compared to +Grid. Experiments demonstrate Starfield's robustness under traffic demand perturbations.