Research On Access Management Technology In Heterogeneous Wireless Network

With the rapid development of modern wireless communication technology,people are already enjoying the scientific and technological achievements brought by the 5thGeneration mobile networks(5G).In the 5G era,due to the diversification of wireless network technologies,a trend of wireless network convergence development will inevitably be formed,and eventually a heterogeneous wireless network will be formed.However,in this kind of network,it is difficult for users to access the optimal network based on the dynamic network environment and their own behavior.In recent years,the rapid development of machine learning has provided new ideas for solving this problem,thereby improving the way of network access,Among them,reinforcement learning technology can learn the optimal strategy through the reward value feedback mechanism.Therefore,this paper studies how to apply reinforcement learning to heterogeneous wireless network access management technology.Achieving the integration between different networks is a complex system engineering.It needs to overcome the technical gap caused by different entry technologies,so there must be a heterogeneous wireless network resource management system to manage the network,so as to better achieve network access control.Based on the existing heterogeneous wireless network resource management model,this paper proposes a resource management system for heterogeneous wireless networks based on WLAN and LTE.This system uses a hierarchical architecture.The entire system is divided into four layers,each one corresponds to a specific network unit,and corresponding functional modules are designed for each level.In addition,a variety of messages are designed for the interaction of various modules in the system.Secondly,this paper proposes a heterogeneous wireless network access algorithm based on Q-learning and Deep Q-Network(DQN)to model the access management problem in the network and define the model state and actions.In order to meet the individual needs of users,different services generated by users are classified,and service types are distinguished for access selection.While fully considering user Quality of Experience(QoE),network Quality of Service(QoS)is also considered.Analytic Hierarchy Process(AHP)is used to define the normalized multi-attribute reward value function that distinguishes different business types,thereby generating feedback on the user's network access behavior.The proposed algorithm can make the heterogeneous wireless network resource management system able to learn the behavior and network status of actual users in a dynamically changing network environment,and ultimately enable users to access the optimal network.Finally,on the NS3 simulation platform,a heterogeneous wireless network resource management simulation model was built,and the function of each module in the model was tested and verified to realize the collection of user information and network resource information.At the same time,the state of the system is defined by binary coding,and then the performance of the access algorithm based on Q-learning under different parameters is compared through simulation,and the influence of different parameters on the algorithm is analyzed.Simulate and compare the access algorithms based on Qlearning,SARSA,use the results of the best performing algorithm to train the Q network in the DQN algorithm under different parameters,and analyze the impact of different parameters on the DQN-based access algorithm.