Reinforcement Learning-Based Handover Strategy for Space-Ground Integration Network with Large-Scale Constellations

The space-ground integration network (SGIN) with large-scale constellations has become an important topic of the next generation mobile communication technology, in which the handover technology between the satellite and the ground is the key technology to ensure the continuity of user service. However, compared with the ground base station's coverage of users, satellites have larger coverage and propagation delay, and large-scale constellations make multiple selectable service satellites above the same user. These phenomena bring great challenges to the handover algorithm. This paper designs a reinforcement learning-based multi-attribute satellite-ground handover strategy (RLMSGHS) for SGIN with large-scale constellations. Firstly, users are clustered with the attributes of location, speed, and bandwidth demand. Then, the handover decision can be made based on the proposed RLMSGHS according to the attributes of received signal strength (RSS), speed, network bandwidth utilization and, handover cost. Finally, the simulation results demonstrate that the heavy decision-making burden caused by the large-scale growth of users in the SGIN is significantly reduced. The multi-attribute handover decision in the SGIN is realized, which reduces the handover demand of users and improves the resource utilization rate of the SGIN.