Satellite Network Multicast Routing Technology Based On Traffic Prediction

As an important part of the space-ground integrated network,the satellite communication network already has large-scale coverage and rich spectrum resources,and can accurately acquire,quickly process and transmit massive business information.It is expected to provide global broadband access and complete data transmission efficiently.However,the topology of the satellite network changes periodically,and the traffic characteristics on the satellite network are relatively more complex.There are many problems of unbalanced distribution of resources and traffic,and different types of services have different requirements for Quality of Service(QoS).Considering the high load traffic and service diversity of the satellite network,multicast routing technology can be used to realize data distribution transmission and information transmission.The thesis is based on satellite traffic forecasting and multicast routing algorithm,and carries out related research.The main work of this thesis is as follows:1)Research on satellite network traffic forecasting based on spatiotemporal characteristics.In order to effectively extract the spatial characteristics of satellite network traffic,which is neglected by traditional forecasting algorithms,and to meet the requirements of forecasting efficiency and accuracy in load balancing scenarios,this thesis models the satellite network traffic forecasting problem,and uses graph neural network and gated recurrent unit to extract Spatial and temporal characteristics of satellite traffic to predict satellite traffic.In addition,in order to focus on key features and avoid forgetting historical information,the graph attention network module is used to optimize the spatial feature extraction effect of the model,and the attention mechanism is used to improve the timing model.Finally,this thesis designs an improved satellite traffic prediction model based on attention mechanism.Through the comparison of experimental simulation results,the model achieves better prediction effect and training efficiency than other models,and the importance of different modules is verified through ablation experiments.2)Research on satellite multicast routing technology based on deep reinforcement learning.According to the characteristics of the dynamic change of the low-orbit satellite network topology,this thesis analyzes the multicast routing strategy optimization problem in the complex and changeable environment state,and studies the satellite multicast routing algorithm with the goal of minimizing the loss of QoS indicators is designed.Through combining deep reinforcement learning and satellite multicast routing technology,the behavior of satellite node path selection is modeled as a Markov decision process,and the satellite traffic prediction result is entered into the model as the node state in the decision process to enhance its understanding of future load conditions.This model can use the neural network in deep reinforcement learning to perceive the satellite network environment,realize load balancing of dynamic satellite networks,and solve routing decision optimization problems constrained by multiple QoS indicators.Through the comparison of experimental simulation results,this algorithm can more reasonably plan the multicast routing strategy compared with other algorithms,and effectively improve the performance of the satellite network system.In addition,in view of the problem that satellite topology snapshot switching causes link interruption and affects data transmission,this thesis proposes a path switching mechanism,which adjusts the transmission path in advance through segment routing technology.Simulation experiments verify that this mechanism can effectively reduce system packet loss rate and delay jitter,and improve the stability of the routing algorithm.