The Modeling Of Aggregate Mobility And Its Applications In Mobile Communication Networks

Mobile communication networks are designed to provide communications servicesto users on the move as infrastructure, but their performance is influenced by the move-ment behavior of users. Currently, the researches about users' movement behavior aremainly focused on individual mobility, causing the lack of in-depth exploration of macro-scopic overall movement behavior of large crowds. This thesis is mainly focused on ag-gregate mobility, based on original data from operating cellular networks. We establishedmodels of aggregate mobility to explore its underlying rules. According to the rules, wedesigned the related protocols to improve the performance of the networks.First, a PCA-based model of aggregate mobility is proposed in order to exploreits space-time structure and evolutional pattern. We applied PCA to original data fromtemporal and spatial perspectives, and discovered three moving patterns: deterministicpattern, spike pattern, and noise pattern. We also discovered its low intrinsic dimension-ality and transient temporal stability, which have applications in forecasting and anomalydetection of aggregate mobility.Second,a local field-based model is proposed in order to describe the local detailsof aggregate mobility,based on the PCA-based model. The model is established fromthe perspective of field. The main problem here is to recover the field itself from theflux of the field under the assumption of keep filed, and it is solved by introducing Po′lyavector field to link the integration of complex function and plane vector field.Third, an energy saving scheme is proposed to solve the problem of high energy con-sumption of base stations, based on the space-time structure of trafc loads. We modeledtrafc loads by applying PCA, according to the space-time relevance between aggregatemobility and trafc loads. We proposed an optimization algorithm of reallocating theexpected trafc loads between cells,in order to minimize the number of active basestations for energy saving. Simulation results show that our scheme has a comparableperformance in energy saving and avoids excess cost of communications and computing.Last, an algorithm is proposed to predict two critical parameters in dynamic locationarea planning. Based on the PCA-based model and the local field-based model estab-lished, we proposed direct and indirect approaches to predict the cost of location updateand paging, which are related with aggregate mobility in dynamic location area planning. Simulation results show that our method has a better performance than common methods.The study about the modeling and applications of aggregate mobility in this thesiswill promote a deep understanding of the movement behavior of large crowds, and theperformance improvement and application development in mobile communication net-works.