Research Of Routing Technology Based On Machine Learning In Integrated Satellite-Terrestrial Information Network

ISTN(Integrated Satellite-Terrestrial information Network)is the integration of heterogeneous networks such as space,air and ground network.With the continuous exploration of space and expansion of Internet applications demand,the integration of satellite and ground will become the trend of the development of the communication network.A certain extent on the study of integration of satellite and ground network represents the forefront of the information technology development level.Due to the wide coverage area,quick speed and the variation of the characteristic parameter of networks,links of low quality,and multiple network protocols running at the same time,the integrated network of satellite and ground has great difficulties in network architecture design,heterogeneous network management,resource allocation,routing and so on.Thanks to the quick development of software-defined network technology in recent years,SDN effectively improves the efficiency of network deployment based on the separation of control plane and data plane.Meanwhile,with the wide application of artificial intelligence and machine learning in solving engineering problems,it also provides a new way of thinking for the research of network architecture and routing technology.The main work of this dissertation consists of the following three points: Firstly,Due to the heterogeneity and the continuous expansion of the scale of the network,the existing network architecture can no longer achieve flexible and efficient network deployment,so this dissertation designed an architecture based on SDN in the ISTN.Based on the idea of separating control plane from data plane in traditional SDN network,we designed a controller for centralized management and control of the global network.And we elaborate the implementation details and deployment scheme of the controller in the following.Secondly,using OPNET and STK simulation software,we set up the integration of satellite and ground network simulation platform based on SDN architecture.And a topology control mechanism suitable for this network scenario is proposed.From south to the principle of interface protocols Open Flow,we design a kind of communication protocol which is applicable to this scenario,then realize the ground terminal controller and satellite node modeling,finally we verified the rationality of the proposed integrated network architecture through the simulation.Thirdly,this dissertation proposes an integrated adaptive routing algorithm ISTN-QR(Integrated Satellite Terrestrial Information Network based Q-Learning Routing Alogorithm)which based on Q-Learning,a branch of machine learning algorithm.Thanks to SDN controller in the centralized control of network architecture,the performance of the proposed algorithm has better performance than that of Q-Learning algorithm on the routing convergence time.At the same time,combining ISTN space-based network in the operation of the satellite orbit characteristics,we improve the way of routing computation of Q value in the process of updating and routing strategy.The simulation results show that ISTN-QR routing algorithm has higher packet delivery rate under high and low traffic loads,and also has significant performance improvement in end-to-end transmission delay and delay jitter,compared with FEQ-Routing and traditional SDN routing algorithm Dijkstra.Therefore,ISTN-QR algorithm has better reliability and stability on the whole.