The Research On Traffic Grooming Technology In 5G Multiple Air Interface

With the rapid development of mobile communication technologies,a variety of new radio access technologies(RAT)and new traffics are emerging.Therefore,there will be more than 2 networks(such as 5G,4G and 3G)simultaneously cover in the same hotspot area with a high possibility.Therefore,the next-generation communication system 5G should be a heterogeneous wireless network which can organically combine the existing networks and the future networks.In the case of 5G fusion networks,it is urgent to solve the problem that how to reasonably groom the traffics to the networks.In order to solve the problem of multi-traffic grooming in the future 5G heterogeneous networks scenario,this thesis focuses on the network selection technology in the traffic grooming technology.As a technology of traffic grooming in access network side,network selection technology can allocate different traffics to different access networks intelligently.It can effectively maximize the network utilization and is the first step in the traffic grooming process.The main contributions of this thesis are as follows:Firstly,this thesis analyzes the types of traffics that may occur in 5G new scenario: transportation traffic,industrial automation and utility traffic,health traffic,virtual reality(VR)/ augmented reality(AR)traffic and smart city traffic.And the quality of service(QoS)requirements of these traffics are also analyzed.The analysis of traffic type is the basis of traffic grooming theory.Aiming at the network selection in traffic grooming,this thesis proposes a network selection method based on bipartite graph multiple matching.This method analyzes the network selection problem from the perspective of graph theory.And the minimum cost maximum flow algorithm is used to get the optimal matching result.Compared with the comparison scheme,the proposed scheme can achieve the balancing of users,improve the performance of network balancing and reduce the overhead of network accessing.In order to analyze the resource competition caused by network selection,this thesis proposes a modified network selection method based on evolutionary game.This method not only improves the reward function used in previous works,but also extends the single traffic scenario to multi-traffic scenario.The simulation results show that the proposed scheme can effectively balance the network load and reduce the switching overhead.In order to further solve the signaling overhead caused by centralized processing in evolutionary game scheme,2 network selection methods based on reinforcement learning(RL)are proposed.They can achieve the balancing of users,improve the performance of network balancing and reduce the total system power consumption.