Low-Carbon Federated Multiagent-DRL Enhanced Network Slicing for Satellite Direct-to-Device Communications

In the evolving landscape of global telecommunications, this study proposes the integration of Satellite direct-to-device (D2D) communication into a federated multiagent deep reinforcement learning (DRL) framework, enhanced by blockchain technology. This approach aims to optimize network slicing capabilities for Satellite D2D communication through low-carbon sustainable computing techniques. The federated multiagent-DRL framework dynamically manages network slices, ensuring optimal resource allocation and superior Quality of Service (QoS) for direct satellite links within 6G nonterrestrial networks (NTNs). Satellite D2D addresses connectivity challenges in remote areas, reducing latency while improving network reliability and coverage. The blockchain component ensures security and transparency, fostering a decentralized learning environment with strong data integrity. By incorporating low-carbon computing methods, the framework not only reduces energy consumption and carbon emissions but also achieves sustainability goals. Comprehensive simulations show significant improvements in latency, throughput, and network resilience, optimizing resource utilization and paving the way for efficient and robust 6G NTN architectures.