DQN-based Network Selection and Load Balancing for LEO Satellite-Terrestrial Integrated Networks

The future communication system will involve collaboration between satellites and terrestrial stations. UEs that directly communicate with low Earth orbit (LEO) satellites will experience frequent handovers, which necessitate effective mobility management for user terminals within the LEO satellite-terrestrial integrated network. This paper proposes a deep Q-learning network (DQN) controller based on the Markov decision process (MDP) to address this issue. This controller can be utilized for adaptive resource allocation and network selection in satellite-terrestrial integrated networks to optimize communication resource utilization and achieve load balancing.