Research On Mobility Prediction Methods And Resource Management Based On User Mobility

In order to sustainably support the explosive growth of traffic amounts and diversified traffic demand patterns,how to optimize the network control mechanism and improve the utilization of network resources has become a key issue.However,the increasing number of mobile users inevitably exacerbates the dynamic changes in wireless services,making network resource management more difficult.Existing studies have shown that user movement has a certain degree of regularity and predictability.Therefore,how to make mobility prediction and apply these information to the design of the wireless network to improve network resource utilization has become a new research hotspot.Fortunately,in recent years,the development of big data and artificial intelligence(AI)technologies has provided new ways for mobility prediction and network optimization control.Based on deep learning technology,this paper comprehensively studies the trajectory prediction problem,and further uses the prediction results to optimize the mobility management mechanism in cellular networks.Considering different application scenarios,this paper divides the trajectory prediction problem into personalized trajectory prediction problem and multi-user trajectory prediction problem.For the former,this paper proposes a Long Short Term Memory(LSTM)-based personalized trajectory prediction model,where the powerful time-series analysis ability of the LSTM algorithm is utilized to learn the user's movement pattern from his/her complete trajectory data.Experiment results in both cell-level and coordinatelevel prediction scenario demonstrate the effectiveness of the proposed model.For the latter,this paper proposes a Sequence to Sequence(Seq2Seq)-based multi-user prediction model to make multi-step trajectory prediction for multiple users in a specific area.Experimental results on a realistic dataset demonstrate that the proposed model has significant improvements on generalization ability and reduces error-accumulation effect for multi-step prediction.Furthermore,this paper proposes a trajectory prediction-based intelligent dual connectivity mechanism,so as to alleviate the handover problem.Simulation results show that the proposed mechanism can significantly improve the quality of service of mobile users in the handover process while guaranteeing the network energy efficiency.This paper further studies how to design an intelligent resource allocation mechanism in the wireless network slicing scenario,so as to proactively adapt to the dynamic changes of service requests caused by the movements of service subscribers.In this paper,the resource allocation problem is modeled as a Markov decision process.Then the LSTM network is employed to capture the temporal regularity of user movement and the amount of different services requests.Based on this,the optimal bandwidth allocation strategy is learned by the help of reinforcement learning mechanisms.The entire solution can be considered as an end-to-end LSTM-based intelligent decision mechanism.Simulation results show that the proposed mechanism can quickly learn the optimal bandwidth allocation strategy under a dynamic environment.The learned strategy can not only guarantee the quality of service of different service subscribers but also improve spectrum efficiency,thereby significantly improving the system utility.