An Intelligent Area-segmentation Enabled Hybrid Routing Method in Mega-constellations

With the rapid development of very high-speed inter-satellite links, effective and intelligent routing strategy has become a hot topic in the next generation (6G) mega-constellations. However, constrained by the high dynamic time-varying topology and limited on-board processing capability, the existed routing schemes incur excessive control overhead, huge latency, and packet loss rate caused by the congestion. In this paper, an intelligent area-segmentation enabled hybrid routing method is proposed, combining centralized and distributed methods and adapting to the congested area dynamically. Specifically, with software-defined features, only the Low Earth Orbit (LEO) satellites with congested links report the state information to the controllers, consisting of both Geostationary Earth Orbit (GEO) satellites and Ground Computing Center (GCC), to periodically calculate the accurate congested areas using Deep Q- Network (DQN) algorithm. Numerical simulation results indicate that, the proposed method reduces control overhead by orders of magni-tude and decreases the packet loss rate by approximately 40% in mega-constellations, toward the future resilient and scalable 6G communication networks compared with the traditional schemes.