Research Of User Mobility Prediction And Handoff Management Technologies In Hetnets

The topic about Heterogeneous Networks(Het Nets) is one of the most important trends and research interests for the next generation 5G wireless communication system. As an important research topic in Het Nets, the performance of mobility management has a direct effect on the total performance of Het Nets and Qo S of user service. There are many differences among these different kinds of access technologies, such as physical frequency, networking mode, service requirement, service terminal access ability and heterogeneity of operation management. These differences make the mobility management of Het Nets different from that of the traditional homogeneous networks. And there are more challenges in the mobility management of Het Nets. Therefore, mobility management of Het Nets has attracted a lot of attentions in the research community.In Het Nets, efficient mobility prediction in mobility management can make mobility management more intelligent and effective. And handoff management is the most vital technology which has a big influence on mobility management performance of Het Nets. Thus, the two key mobility management technologies for Het Nets will be mainly investigated in this paper.Firstly, network topology, network state parameter analysis and the mobility model establishment of new users are given out, which are the fundamental research of mobility prediction and handoff management. Among them, the single-cell and multi-cell topology descriptions are the basis of the mobility prediction and handoff managements for the two different network scenarios. And the network state parameter analysis is the most important handoff decision-making basis for handoff management in this paper. In addition, this paper brings in new mobility patterns for users, which accords with the real mobility rules of users in Het Nets. The introduction of new mobility patterns not only makes the mobility prediction more realistic, but also enhances the rationality and practicability of mobility management for Het Nets.According to the characteristics of single-cell topology, an improved Markov mobility prediction(IMMP) scheme is proposed based on the classic mobility model. Compared with the other algorithms, the complexity of the new scheme is the lowest and the prediction accuracy is the highest. Moreover, for the complicated multi-cell topology of Het Nets, a multi-class support vector machine(SVM) mobility prediction scheme is presented for predicting the locations of macro mobility users under the new mobility model.The aim of handoff management is to enhance the network ability and profit. Thus, to enhance network capacity and decrease the call blocking rate and handoff dropping rate of new users, an improved Markov decision process(IMDP) vertical handoff scheme suitable for single-cell topology of Het Nets is presented with handoff decision-making factors, i.e., sub-network port state factors. Additionally, for multi-cell topology of Het Nets, we propose Q-learning based multi-cell hybrid handoff scheme considering network state factors and Qo S constraint of users comprehensively. Via the online Q-learning algorithm, horizontal and vertical handoff between different sub-networks can be achieved. On the premise of guaranteeing Qo S of users, the new algorithm enhances the network capacity and utilization.The high requirements of wireless service quality for users result in the fierce competition between Networks. Thus, only the better quality of experience(Qo E) for users can bring the higher profit for the networks. The main factors of Qo E for users include the objective Qo S parameters, service fees and the energy consumption of the terminal. Via the analysis above, a Qo E evaluation scheme is proposed based on random neural network(RNN), through which we can find the relationship between Qo E and Qo S. Then, Qo E-based multi-agent Q-learning handoff strategy is proposed with the aim of maximizing Qo E. The proposed scheme takes service fees and the energy consumption of the terminal into consideration comprehensively, which can enhance the whole network profit.