Machine learning-inspired intelligent optimization for smart radio resource management in satellite communication networks to improve quality of service

Abstract

Satellite communication networks are seeing a significant surge in traffic requirements. Nevertheless, the rise in traffic requirements is inconsistent across the service region because of the unequal distribution of consumers and fluctuations in traffic requirements during the day. Variable payload designs solve this issue by enabling the uneven allocation of payload resources to match the traffic requirement of each beam. Optimization-based Radio Resource Management (ORRM) has its high substantial efficiency demonstrated computational difficulty hinders its real-world deployment. This work explores the structure, execution, and uses of Machine Learning (ML) for resource allocation in satellite systems. The primary emphasis is on two systems: one that offers power, capacity, and beamwidth adaptability and provides temporal flexibility via beam hopping. The research examines and contrasts several ML methods suggested for these structures. The research determines whether training must be done online or offline depending on the features and needs of each ML method. The study analyzes the most suitable system structure and the pros and cons of each strategy.