Rethinking Nearest-Is-Best: Towards Load-Aware Ground Station Assignment for LEO Satellite Networks

Low Earth Orbit (LEO) satellite constellations promise global connectivity with low latency. Yet, their ground segment often becomes a bottleneck because uneven user demand overloads certain Ground-to-Satellite Links (GSLs) while leaving others underutilized. Current operational LEO satellite constellations (e.g., Starlink) primarily route user traffic through the nearest Ground Station (GS) to minimize propagation delay. However, nearest is not always best for load balancing. This routing policy can overload a small set of GSs and their GSLs, creating bottlenecks in the network. In this paper, we propose a load-aware assignment strategy that minimizes the maximum GSL utilization by selectively routing some users to non-nearest GSs when beneficial. By leveraging inter-satellite links, user data is allowed to be forwarded through the satellite mesh to alternate GSs with spare capacity. To preserve service quality, we introduce a latency threshold so that assignment is allowed only when the additional delay remains within a small margin of the nearest-GS baseline. We formulate the problem as a mixed-integer linear program that explicitly minimizes peak GSL load under delay constraints. Evaluation using realistic Starlink constellation parameters shows that the proposed solution significantly reduces average maximum GSL utilization by up to 60.77%, alleviating ground bottlenecks while keeping added latency within acceptable limits.