Capacity Analysis and Robust Topology Design of LEO Satellite Constellation With Node/Link Failure

Low Earth Orbit (LEO) satellite networks are increasingly central to wide-area communication services, yet their capacity is highly sensitive to the inherent vulnerability of satellite and inter-satellite link (ISL) failures. This paper develops an analytical framework for characterizing network capacity under arbitrary failure patterns. Building on bisection cut-set analysis, we formalize capacity bottlenecks through graph partitioning and derive rigorous upper bounds on the network capacity. Within this framework, we further identify and characterize the maximum failure cut set, i.e., a critical subset of ISLs whose removal yields the most severe capacity degradation, and establish its necessity in achieving the worst-case capacity bound. Furthermore, the analytical framework has been shown to be applicable to general LEO networks, and a closed-form expression for capacity degradation has been derived using a 2D torus topology as an example, which takes into account both the number and spatial distribution of the satellite or ISL failure. The results indicate that maximizing the minimum bisection cut set capacity is fundamental to strengthening the intrinsic robustness of the LEO network. This paper provides a theoretical methodology for capacity assessment and offers principled guidelines for topology design in satellite constellations.