Research On Access Selection Algorithm In Heterogeneous Wireless Networks

With the rapid development of wireless communication technology,the diversity of wireless network technologies trends to become increasingly apparent.All kinds of wireless network collaboration complement each other constitute the heterogeneous wireless network,which become the inevitable trend of integration of wireless network development.It's of great importance to choose appropriate access networks to realize heterogeneous wireless network resource management,improve network capacity and quality of user experience.Therefore,this article mainly researches on effective access methods in heterogeneous wireless network.Because of the diversity of traffic type and the more prominent personalized needs of different user,the traffic requirements also become obviously diffierent,and we need to access networks with distinguishing between traffic types.Most of the existing network selection algorithms focus more on Quality of Service(QoS)as the network selection criteria,emphasizing the overall network performance,but ignoring the specific needs of different users.They have not considered the availability and dynamic of network state information,and can't make a valid choice decisions.In view of the above problems,this thesis introduces the theory of Multi-Armed Bandit(MAB)problem,which senses the dynamic changes of the network without interrupting the communication,and matches the optimal network for all kinds of traffic through online-learning.Firstly,the classification of MAB is summarized and the distribution characteristics of such models are verified,and the influences of different scales and variances of reward on decision is realized,then the selection of network scene parameters is realized.Secondly,considering the network QoS and user Quality of Experience(QoE),the Analytic Hierarchy Process(AHP)is used to define the normalized multi-attribute reward function which distinguishes the service type to feedback the influence of the access on the users.Based on the MAB selection algorithm,a traffic-aware network selection algorithm is proposed,which transforms the single decision goal in the MAB model into multi-objective decision-making.This kind of network selection algorithm considers the QoS requirement while focusing on the user's needs,and the self-learning characteristic of the MAB model is used to perceive the dynamic changes of the network,access the optimal network for all kinds of services,and make better use of existing information to speed up the convergence rate.Finally,the continuous-time MAB(CTMAB)model is studied,and an access selection algorithm considering the vertical handoff is proposed.CTMAB takes into account the duration of the reward,introducing of the reward rate as a decision feedback;Using this CTMAB feature,the reward rate is defined as a rate of reward and handover cost;The traffic in its duration is divided into a number of decision slots,in each slot to determine the network switch.The algorithm dynamically learns to realize the access selection,and realizes the vertical switching of the long service.