Connectivity Analysis Of Opportunistic Mobile Networks And Its Application:Time Evolving Perspective

With the popularization of mobile smart devices, Opportunistic Mobile Networks (OppNets), which are easy to deploy, free to use and without the constraints of location in practice, have drawn the interest of many researchers in the area of network and communication. Now, OppNets are gradually becoming effective supplements of infrastructure network. Especially, information technology, marked by Mobile Internet and Internet of Things, is triggering a new round of technological revolution worldwide recently. It also brings unprecedented opportunities to the research and application of OppNet.However, owing to frequent node mobility and communication link changes, OppNets are characterized by dynamic and time-evolving connectivity. Few theoretical tools today can be used to analyze this characterization, which enable us to have a poor understanding of OppNets: Researchers still can’t obtain the communication performance bounds of such networks, and thus can not evaluate the extent to which such networks meet application requirements. Moreover, due to the fact that many critical applications in OppNets rely on great insights to the essence of dynamic connectivity, extensive applications of OppNets remain difficult up to today. For this reason, on the basis of the state-of-the-art literature review, we conduct a more comprehensive research on dynamic connectivity of OppNets in this dissertation. Major research contents and contributions are summarized as follows:(1) We propose a model-Critical Journey Evolving Graphs (CJEGs) to uncover the evolution of end-to-end connectivity of OppNets. For the first time, foremost journeys between any two nodes can be calculated out in continuous time from CJEGs; we design an efficient algorithm CJEG-PERST to derive CJEGs online; furthermore, a simulator CjegSim, which implements CJEG-PERST, is developed. We exploit CjegSim to analyze the basic dynamic connectivity of several real and synthetic OppNets.(2) We continue to explore the utilizations of CJEGs in-depth:Based on CJEGs, some time evolution-oriented metrics to reveal the dynamic connectivity of OppNets are designed; based on CJEGs, new temporal centralities are proposed, which can identify key nodes effectively with low computational cost; new approaches to derive temporal reachability graphs and affine graphs from CJEGs are introduced; we also exploit CJEGs to investigate the communication performance bounds of OppNets in specific applications, and design some strategies to solve practical problems in critical applications of OppNets.(3) To cope with distributed application and lightweight computing environment on mobile nodes, we investigate connectivity of OppNets by exploiting community and centrality concept in complex networks. In OppNets which depends on people’s mobility, two key points need to be considered when we design application strategies:First, contacts often appear periodically; second, the duration of the observation time span plays an important role in the estimation of connectivity probability. Based on these two points, TTL (time to live) community and TTL centrality are proposed to improve the prediction of connectivity probability during the message’s lifetime. Furthermore, a novel social-aware routing algorithm in OppNets, PerEvo, is developed. Extensive trace-driven simulation results show that PerEvo achieves higher performance than the existing social-based forwarding schemes.(4) Since the community can be used as an indirect way to estimate the reachability between nodes, we propose a novel evolutionary community detection called EFOCS; On the basis of EFOCS, a new reachability-related centrality named OR_CEN is designed. Content distribution experiments show that OR_CEN can estimate the influence of nodes effectively. Furthermore, several approaches to choose initial target sets according to the influence of nodes are designed, and then applications of these approaches in mobile data offloading are explored. Analysis and simulation results demonstrate that CBS_OR, the approach to choose initial target sets according to OR_CEN, has more outstanding offloading capability than other approaches. In particular, OR_CEN is easy to implement. Therefore, CBS_OR can be regarded as an ideal lightweight offloading strategy.To sum up, in our research, dynamic and time-evolving connectivity of OppNets are explored systematically in a fine-grained and a coarse-grained analysis way respectively. In the fine-grained analysis, nodes are taken as analysis units, and then the end-to-end connectivity are calculated directly; while in the coarse-grained analysis, communities are taken as analysis units, and then the end-to-end connectivity are predicted indirectly. By our research, theories with regard to connectivity modeling of dynamic network are further enriched, and thus the communication performance bounds of OppNets can be estimated more accurately, which provides powerful support for extensive applications of OppNets. Moreover, a series of methods to characterize the connectivity are established, and some strategies with good performance in routing and mobile data offloading are developed. By exploiting Opportunistic Network Environment (ONE) and CjegSim, we conduct a great amount of empirical studies. Experiment results show that the model and the approaches proposed in this dissertation can capture critical characterization concerning dynamic connectivity of OppNets. Based on the models and approaches proposed in this dissertation, many application stragtegies are designed and proven to have improved the application performance.