Research On The Routing Algorithm Of Space-Ground Integrated Network Based On Reinforcement Learning

Space-Ground Integrated Network(SGIN)has the advantages of large spatial span,wide coverage range,high robustness and strong invulnerability,which can effectively address the problem of surging demand for communication services,thus has become an important research branch in the Internet field with broad application prospects.Meanwhile,SGIN has the characteristics of large network scale,complex and heterogeneous network structure,dynamic change in network topology,and air-ground communication severely affected by weather,which also makes this research field full of challenges.Reinforcement learning provides powerful empirical and feature learning capabilities,using reinforcement learning to solve the SGIN routing problem makes it possible to optimize the SGIN routing algorithm and improve the transmission performance and efficiency of the network.This thesis investigates the routing algorithm for SGIN,analyzes the influencing factors of air-ground communication and proposes an attenuation prediction method of the SGIN air-ground communication link,then designs a routing optimization algorithm based on deep reinforcement learning for SGIN.The main work and results of the thesis are as follows:1.The impact of weather on air-ground communication is analyzed,and a deep learning-based air-ground communication link attenuation prediction model is proposed,which predicts the rainfall level by multi-dimensional weather-related features and accurately estimates the communication link path attenuation between satellites and ground by combining communication frequency,communication distance and other parameters,providing a basis for the SGIN route optimization problem in the face of severe weather.2.The SGIN routing algorithm is studied,and an SGIN routing optimization method based on the DDPG algorithm is proposed to make dynamic routing decisions based on real-time link state information,and long short-term memory network(LSTM)is utilized in the network model to improve the algorithm’s ability to perceive the temporal characteristics of network data flows.3.An air-ground integrated network structure is built based on the SGIN simulation platform OMNET++,which realizes the simulation of networking,data transmission and routing.The experimental results verify the effectiveness of the proposed routing optimization algorithm,which is significantly better than the traditional OSPF routing algorithm in terms of throughput,average link delay and packet loss rate of the network under different traffic intensities.