Scalable Satellite Handover Management in Nonterrestrial Networks: A Distributed MADQL Approach

Satellite communications provide means for extending next-generation communication technology to areas beyond terrestrial network coverage. Multiple Low Earth Orbit (LEO) satellites have been deployed in constellations to offer ground users direct Internet connection at any time, place, and condition. However, integrating Non-Terrestrial Networks (NTNs) into terrestrial communication systems presents several challenges, including the management of handover strategies due to the rapid satellite movement. Even if previous studies have explored various approaches to optimize handovers in NTNs, they have often overlooked critical factors, such as user-specific data throughput requirements, limited satellite energy, and the need for dynamic adaptation to real-time network conditions. To address these gaps, this paper proposes a Scalable Multi-Agent Satellite Handover (SMASH) framework based on Distributed Multi-Agent Deep Q-Learning for optimized handover decisions and dynamic satellite selection. SMASH aims to ensure seamless connectivity while effectively managing satellite resources to meet diverse demands. It features an adaptive resource allocation strategy driven by user demands and network conditions, thereby ensuring application-specific Quality of Service requirements. We validate and evaluate SMASH using a satellite network simulator, conducting a comprehensive sensitivity analysis and benchmarking its performance against existing approaches in the literature. The proposed handover technique significantly enhances NTN communication performance by reducing the average number of handovers and optimizing satellite resource allocation, thereby preventing user blocking. The SMASH framework ensures continuous service delivery by dynamically adapting to fluctuations in both user demand and satellite resource availability.