Deep-Learning-Based Resource Allocation for Multi-Band Communications in CubeSat Networks

Abstract

CubeSats, a type of miniaturized satellites with the benefits of low cost and short deployment cycle, are envisioned as a promising solution for future satellite communication networks. Currently, CubeSats communicate only with ground stations under limited spectrum resources and at low data rates, whereas with growing launches of CubeSats and more diverse services expected every year, novel communication techniques and resource allocation schemes should be investigated. In this paper, a multi-objective resource allocation strategy is designed based on deep learning algorithms for autonomous operation in CubeSats across millimeter wave (60–300 GHz) and Terahertz band (300 GHz-1 THz) frequencies with the utilization of reconfigurable plasmonic reflectarrays. Simulation results demonstrate the inter-satellite links can achieve multi-gigabits-per-second throughput and ground-to-satellite links with more than 10 times of capacity enhancements in realistic channel conditions.