Handover Optimization for Aeronautical Communications in LEO Satellite Networks: A Deep Reinforcement Approach

As key fundamental components of the space-air-ground integrated network (SAGIN), low earth orbit (LEO) satellite networks provide critical support for aeronautical communications in remote regions such as polar areas and oceans. However, the high-speed mobility of LEO satellites and the expanding scale of constellations lead to frequent handovers, resulting in severe communication overhead and propagation delay in data transmission. Therefore, investigating handover optimization strategies is essential for aeronautical communications. Several attributes, including achievable data rate, idle channels, service time, and propagation delay, are taken into account to make handover decisions. We propose a strategy to jointly optimize handover frequency and quality of service (QoS). Considering that the actions of other aircraft bring uncertainties into the decision-making process, we propose a novel deep reinforcement learning (DRL)—distributionally robust soft actor-critic algorithm, which can guarantee worst-case performance, ensuring the operational safety of the aircraft. Simulation results show that the proposed handover optimization scheme effectively reduces handover frequency while maintaining a high success rate. Furthermore, the results demonstrate that the proposed algorithm outperforms conventional deep reinforcement learning algorithms in terms of learning efficiency and robustness.