Traffic-Density Adaptive Service-Area Partitioning Algorithm in LEO Satellite Networks

Abstract

Low Earth orbit (LEO) satellite networks have become a key component of space-air-ground integrated networks due to their wide coverage, low latency, and flexible networking capabilities. To address uneven ground traffic distribution, hotspot migration, and the limited adaptability of fixed-granularity service regions in large-scale constellations, this paper proposes a traffic-density-adaptive dynamic service-region partitioning method. First, a non-uniform ground traffic grid is constructed, visible satellites are screened for each grid cell, and the primary serving satellite is selected according to the access elevation angle and service retention threshold. Ground traffic is then aggregated into equivalent satellite loads based on the primary-service association. Next, local traffic density is evaluated with inter-satellitelink topology, and the target region size and diameter are adaptively determined. Connected service regions are generated and maintained through constrained region growing, hotspot splitting, sparse-region merging, and boundary migration. Simulation results show that, compared with fixed orbital-grid partitioning, latitude-topology-based partitioning, and ADWCA, the proposed method improves Jain's load-balancing index by 0.176, 0.090, and 0.243, respectively, while enabling adaptive adjustment of serviceregion granularity, boundaries, and structure as traffic hotspots evolve.