LEO Satellite Network Routing Algorithm Based on Reinforcement Learning
Xiaoting Wang, Zhiqi Dai, Xu Zhao
Abstract
Satellite networking is a network composed of satellite connections, through which global coverage can be achieved to meet the communication requirements of users in remote areas. Low-orbit (LEO) satellites have the advantages of small transmission delay and global coverage, and have always been a concern hot spot. Due to the fast running speed of LEO satellites and its frequently changing satellite topology, it is difficult to directly apply routing protocols applied on the ground to LEO satellite networks. Therefore, it is necessary to design a suitable routing algorithm to make the LEO satellite network run smoothly. This paper first models the operating environment of a LEO satellite network. Then it proposes a satellite routing algorithm based on reinforcement learning according to the operating characteristics of LEO satellites and improves the traditional reinforcement learning algorithm. Simulation results show that the improved algorithm has better convergence time and results than the previous algorithm.