Resource Allocation and Load Balancing for Beam Hopping Scheduling in Satellite-Terrestrial Communications: A Cooperative Satellite Approach

Abstract

Satellite-terrestrial communications based on mega low-Earth orbit (LEO) constellations enable extensive coverage and high data rates. However, the communication performance is significantly impacted by the non-uniform traffic distribution and the substantial interference caused by dense satellites. Therefore, in this paper, a multi-satellite cooperation architecture for satellite-terrestrial communications is proposed in LEO satellite constellations. Specifically, beam hopping (BH) and resource allocation enable a flexible solution for the non-uniform geographical distribution of communications, and are optimized to improve communication performance and avoid both intra- and inter-satellite interference in satellite-terrestrial communications. Moreover, load balancing via inter-satellite link (ISL) is implemented to further enhance the communication performance. Consequently, the overall problem for maximizing the network throughput and ensuring the latency metric is formulated, and then decomposed into three sub-problems: deep reinforcement learning (DRL) based BH scheduling, resource allocation with multi-satellite cooperation, and load balancing via ISLs. Specifically, DRL is utilized to determine the real-time BH pattern, and the resource allocation among beams is implemented by the majorization-minimization algorithm. Furthermore, for varying input traffic loads, different objectives for load balancing are established and solved by quadratic transformation and hybrid block successive approximation algorithm. Simulation results demonstrate that the proposed method outperforms other existing methods. Meanwhile, it obtains an 18.45% improvement in throughput compared with the benchmark without optimization, while the latency metric for satellite-terrestrial communications is also reduced by the proposed method.