Access Network Selection Technology In Ultra-dense Wireless Networks

As one of the key technologies of the fifth generation mobile communication,ultra-dense networks increase the spatial resources utilization and thus improve the network capacity in hotspots by densely deploying infrastructures.However,as the density of base stations increases,the user lies the coverage of multiple cells,which leads to the greater probability of multiple users accessing the same cell and makes the network more susceptible to overload.In addition,the load in each cell greatly varies with the cell radius,which makes the general prediction method hard to provide useful information for access network selection.Meanwhile,the reduction of the cell radius will cause users to frequently switch between multiple base stations,which makes the regularity of data switching between cells is poor,and makes the problem of resource reservation becoming a challenge.The article has conducted in depth research and study on the above issues.The main work and contributions are listed as follow:First of all,the Ridge Regression and Random Forest model in machine learning are combined to predict the number of users in ultra-dense network,which increases the prediction accuracy.Depend on this result,based on the difference between the predicted value and the maximum number of bearers of the network,and considering the subjective user preferences and object network properties,a new utility function is designed,and then a network access algorithm based on load predicting is proposed,which improves the probability of network access users,and achieves network load balancing.Secondly,aiming at the problem of overloading the target cell when multi-user access in a dense network environment,the competition between users and the network is modeled as a cooperative game problem,and a new revenue function is designed,which comprehensively considers the impact of the performance of network and the user experience on network access selection.Based on this,a network access algorithm based on analytic hierarchy process and alliance game is proposed to improve the comprehensive utility value of the network.Finally,aiming at the difficulty of predicting the number of handovers due to the high temporal and dynamic characteristics of the handover between users in ultra-dense network,based on the Gradient Boosting Decision Tree model,a multi-scheme fusion prediction algorithm is proposed.From the number of handovers themselves,the transition probability,the ratio of handovers to the number of users,and fuses the corresponding prediction result,by mining the regularity of different types of data,the prediction accuracy is improved.Furthermore,according to the predicted value of the handover number,corresponding resource reservation is performed,and the utilization of network resource is effectively improved.