| With the rapid development and maturity of 5G technology,more and more people enjoy the high-speed,low-latency experience brought by 5G.However,due to the coverage and energy consumption of 5G base stations,5G can only cover very few areas.Relatively speaking,a few high-orbit satellite networks can more easily cover most areas of the earth and provide users with global services,but the delay is high and the signal is greatly affected by the environment.As an effective supplement to the terrestrial cellular network,the low-orbit satellite network has a large coverage area,the launch cost is continuously reduced,and the bandwidth and delay are greatly improved compared with the high-orbit satellite.However,it has strong mobility and requires more global coverage.number of satellites.Therefore,it is necessary to study how to effectively combine terrestrial and satellite networks to achieve a low-latency network with global coverage at a lower cost.In the above background,this paper comprehensively considers the characteristics of small ground network coverage,high delay,and fast mobile speed of satellite networks,and carries out research on core network architecture design,resource allocation,and service migration under the background of satellite-ground fusion networks.The main research contents are:1.Aiming at the network environment of star-earth integration,this paper proposes a user-centered core network architecture and builds a starearth integration network simulation platform and a user-specific core network management platform based on the open-source software open5gs.The user-specific core network consists of improved AMF,SMF,UPF,and newly added GPF(gNB Proxy Function),CMF(CN Manage Function),and LUD(Local User Data).This lightweight core network can be quickly deployed,migrated,and destroyed.Provide users with low-latency services by deploying on the edge side and the ground cloud,and connecting to the cloud through satellite links to provide users with services with wider coverage;by automatically reclaiming resources on the user-specific core network,it greatly reduces the user’s idle time.resource consumption.Finally,based on the core network management platform,the user’s active selection and dynamic destruction of the core network are verified;UERANSIM is used to simulate users,and it is proved that the user-only core network can effectively reduce resource consumption compared with the traditional core network;finally,through cloud deployment of the core network,The comparison of UPF and UPF sinking methods proves that the user-specific core network can reduce the user’s connection delay and transmission delay in the ground network and satellite network.2.Aiming at the problem of multi-user dedicated core network resource allocation in the satellite-ground fusion network,an improved genetic algorithm based on kinship is proposed,and a resource allocation mechanism for the user-specific core network is designed according to the algorithm.This mechanism takes into account the link characteristics and node resource constraints and designs the default core network and userfriendliness mechanism.Users with low friendliness have a higher probability of using the dedicated core network,and use the improved genetic algorithm based on kinship to solve the problem Model.The simulation experiment proves that compared with the traditional genetic algorithm,the genetic algorithm based on kinship has a faster iteration rate and lower loss value in solving multi-user resource allocation problems;compared with the random selection algorithm and traditional genetic algorithm,this algorithm can reduce user The average delay and reduce the cost per unit of user experience(Quality of Experience,QoE)(Cost per QoE,CPQ).3.Aiming at the high-speed mobile low-orbit satellite environment in the satellite-ground fusion network,this paper designs a user-specific core network migration solution based on DRQN(Deep Recurrent Q-Network).In the low-orbit satellite communication environment,link delay and bandwidth will fluctuate regularly with the movement of low-orbit satellites,and users will generate large delays in the process of continuous switching and reconnection.The DRQN-based core network migration solution proposed in this paper uses the DRQN model to predict the delay and bandwidth characteristics of different links and fit the Q value of different links,and deploy user-specific core networks(AMF,SMF,UPF,and LUD)in advance to reduce the delay of user switching and avoid frequent core network switching.The simulation experiment proves that in the satellite-earth fusion network,the core network migration algorithm based on DRQN can reduce the average switching frequency and CPQ of the core network for users in a period compared with the traditional DQN algorithm and the greedy algorithm. |