Research And Implemrntation Of Routing Algorithm Based On Cooperative Traffic Prediction In Satellite-Terrestrial Network

Benefiting from the advantages of the Low Earth Orbit(LEO)satellite network with low propagation loss,large orders of magnitude,less restricted by the geographical environment,and providing seamless global coverage,the satellite-terrestrial network that closely integrates the satellite network and the terrestrial network has been proposed and received much attention in industry and academia.With the continuous increase of services that can be carried by the satellite-terrestrial network,the construction of a real-time,efficient,and balanced routing algorithm is directly related to the network Quality of Service(QoS),which is an inevitable requirement for the development of the satellite-terrestrial network.However,the topology of the LEO satellite network changes frequently,and the uneven distribution of resources and traffic bring challenges to the design of the routing algorithm in the satellite-terrestrial network.At present,most of the existing studies consider terrestrial network resources and satellite network resources separately,and often combine the respective routing results of the two networks,which is difficult to meet the end-to-end QoS requirements of users.Moreover,the traffic control measures are passively implemented when the network is already congested,and the historical data of satellite traffic is not fully utilized.Aiming at the shortcomings of current satellite-terrestrial network routing algorithms,this thesis proposes a load balancing and QoS routing algorithm based on traffic prediction(TP-LBQR).First,by predicting the traffic of the satellite node at the next moment,reflecting the satellite load,so as to select a more reliable next hop.Because traditional traffic prediction algorithms are lacking in prediction accuracy and efficiency,this thesis proposes a deep learning traffic prediction algorithm based on Stacked Denoising Autoencoder(SDAE).In addition,considering the limited computing and storage capabilities of satellite nodes,a cooperative traffic prediction model(CTPM)for satellite-ground station is proposed to minimize the delay required for prediction.Secondly,the introduction of Software Defined Network(SDN)into the satellite-terrestrial network can effectively solve the management problem of heterogeneous networks,making end-to-end routing possible.Therefore,this thesis establishes an SDN-based satellite-terrestrial network scenario,and defines the link cost function according to the characteristics of the LEO satellites,uses the traffic prediction results to define the load weight factor,and then uses the ant colony algorithm to solve the route to obtain a load balancing and QoS end-to-end path.Finally,a simulation system is designed and implemented based on TP-LBQR routing algorithm.Firstly,the system requirements are analyzed in detail,and the system architecture and database are designed.On this basis,the detailed design of each module is introduced,including network topology display module,traffic prediction module and routing module.The front-end uses Vis.js to draw a visualized topology map,uses jQuery to achieve front-end and back-end ajax interaction,and uses JSP and CSS/HTML to provide visual interface and user interaction functions.The back-end development uses the SSM framework,and the database uses MySQL.Finally,the test situation of the system is introduced.Through functional test and performance test,the stability and availability of the system are guaranteed,which shows that the system can help operation and maintenance personnel to scientifically plan routes in the satellite-terrestrial network.