Optimization for LEO Satellite-Ground Integrated Networks via Deep Reinforcement Learning

Witnessing an emerging communication platform of low-altitude earth orbit (LEO) satellites (SATs), integrating LEO SAT networks and existing terrestrial networks is getting attention. In this paper, we study the optimization of space-ground-integrated networks. Towards maximizing the communication efficiency, using a ground terminal as a relay terminal for the orbiting LEO SAT networks should be considered and optimized; however, that is tricky due to the time-varying network topology and a huge number of possible control actions. To tackle the challenge, a deep reinforcement learning method is used. Simulation results validate that SAT-ground relay (GR)-integrated scheme achieves 2.9x higher end-to-end sum throughput compared to a benchmark scheme with only-SAT scheme.