TY  - JOUR
T1  - Handover Minimization Scheme Using Multi-Agent Deep Reinforcement Learning in Multi-Beam Low Earth Orbit Satellites
AU  - Lee, Chungnyeong 
AU  - Kim, Taehoon 
AU  - Bang, Inkyu 
AU  - Chae, Seong Ho 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2025
DA  - 2025/1/1
DO  - 10.7840/kics.2025.50.8.1196
KW  - Low earth orbit satellite
KW  - handover strategy
KW  - multi-beam
KW  - multi-agent deep reinforcement learning
AB  - In this paper, we propose a Multi-Agent Proximal Policy Optimization (MAPPO)-based handover strategy for multi-beam Low Earth Orbit (LEO) satellite networks, employing the Centralized Training and Decentralized Execution (CTDE) approach of Multi-Agent Deep Reinforcement Learning (MADRL). The proposed strategy aims to minimize the number of handovers and maximize throughput by considering the cost differences between inter-beam and inter-satellite handovers, user quality of service (QoS) constraints, and load balancing.
Each user independently makes handover decisions based on local information (e.g., load and channel conditions within the coverage area), allowing for prompt immediately adaptation to the dynamic and complex environment of multi-beam LEO satellite networks. Simulation results indicate that the proposed algorithm reduces the number of handovers by 39.1% to 75.53% and improves throughput by 14.6% to 157.7% compared to benchmark handover algorithms, thereby objectively demonstrating the superior performance of the proposed approach.