Intelligent Cloud-Edge Collaborations for Energy-Efficient User Association and Power Allocation in Space-Air-Ground Integrated Networks

In space-air-ground integrated networks (SAGINs), the global energy efficiency (GEE) is a crucial metric for balancing the network throughput and energy consumption, and the maximization of GEE requires the optimizations of both user association and power allocation. Most existing methods optimize user association and power allocation separately or successively, relying on instantaneous non-local channel state information (CSI) exchanges. Nevertheless, both the separate and successive methods may fail to find the jointly optimal solution, and acquiring the instantaneous non-local CSI across the SAGINs is challenging due to the long communication distances between the access points (APs) and users. To address these issues, we leverage cloud-edge collaborations and propose an online delayed-interaction collaborative-learning independent-decision multi-agent DRL (DICLID-MADRL) algorithm. With the proposed algorithm, each AP can independently select users and configure transmit power with only local information to enhance GEE. Simulation results demonstrate that the proposed algorithm achieves a higher GEE with reduced time complexity compared to the state of the arts.