Analysis of Graph-based User Scheduling For Ka-Band LEO NTN Systems

This work focuses on the next generations of multi-beam (MB) Low Earth Orbit (LEO) based NonTerrestrial Network (NTN) architectures that exploit full spectrum reuse schemes and effective interference management techniques by employing user scheduling before precoding at the transmit side. Given a high density of user terminals (UTs) on-ground as compared to on-board LEO satellite antennas it is extremely challenging to schedule users. To improve the performance, optimal user scheduling is necessary. In this paper, we present the analysis of graph-based user scheduling algorithms for Ka-band, with a focus on feed space (FS), downlink (DL), and Frequency Division Duplex (FDD) mode. We propose both a distance-based iterative graph-based approach and a distance-based implementation of the Multiple Antenna Downlink Orthogonal User Clustering (MADOC) algorithm. Instead of considering the Coefficient of Correlation (CoC) as a dissimilarity measure to organize users into clusters, we consider the great circle distance between the users as a dissimilarity metric. Once users are organized into clusters, they are served by the LEO satellite using Space Division Multiplexing (SDM) through Minimum Mean Square Error (MMSE) beamforming applied per cluster. Subsequently, these clusters are allocated to different time slots using Time Division Multiplexing (TDM). To validate the effectiveness of our presented approach, we conduct a comparative analysis with the original Channel State Information (CSI)-based MADOC algorithm which is based on the CoC. The results are presented in terms of the achievable per-user throughput (Mbps), and signal-to-noise plus interference ratio (SINR). By considering user scheduling and beamforming together, this study provides valuable insights into the effective use of resources in LEO MB-NTN systems.