Research On Caching And Access Strategies For 5G Satellite-terrestrial Integrated Network

The integration of terrestrial 5G network and satellite communication network,can form a global mobile communication network that covers ocean,land and the air,which is the trend of future communication development.satellite-terrestrial integrated network is composed of many kinds of networks.It has the characteristics of complex network structure and diverse network resources,which makes it difficult to manage network resources.Due to the explosive growth of users and business traffic as well as the emergence of new types of business,new caching and access strategies are required in communication system,to cope with the high latency of satellite communications and to meet user requirements.Therefore,the thesis studies the caching strategy and access strategy of future satellite-terrestrial integrated network with deep reinforcement learning.Firstly,in the two-level cache scenario of satellite and ground station,this paper constructs a cache strategy optimization problem aiming at the total file acquisition delay.Because of the unknown popularity of files,it is difficult to get good performance by using traditional algorithm.Therefore,combined with deep reinforcement learning,this paper proposes a passive caching strategy based on asynchronous advantage actor critical(A3C).At the same time,short-term file requests and long-term file requests are introduced into the state space to deal with the dynamic change of popularity caused by high-speed satellite movement.Experimental results show that compared with the traditional cache algorithm,the average file acquisition delay of the proposed cache algorithm can be reduced by about 15%.Secondly,the thesis studies the satellite access strategy in the scenario of satellite ground two-level cache network.The traditional access strategy is difficult to make efficient access satellite selection according to the characteristics of satellite-terrestrial integrated network.In addition,the high-speed movement of LEO satellites leads to the dynamic change of network topology,which makes the algorithm design difficult.To solve these problems,this paper proposes a satellite access strategy based on Deep Q-learning Network(DQN).The strategy makes access decision which based on satellite cache state,link load and satellite-ground distance information to minimize the total time delay of system file acquisition.Simulation results show that the average file acquisition delay of the proposed access algorithm can be reduced by about 25%compared with the traditional satellite access algorithm.