Research On Data Delivery In Opportunistic Networks

The development of communication technology is leading to a world replete with small wireless devices with a variety of sensing and computing functions. Opportunistic network exploits the opportunistic communication between pairs of devices to share each other’s content, resources, and services. A typical scenario of opportunistic networks is that devices are carried by humans. In such scenario, the movements of the communicating devices strongly rely on the mobility patterns of their owners.Data delivery in opportunistic networks relies on nodes’mobility to route messages to destinations as they are encountered. Due to time-varying links and intermittent connectivity, routing is a challenging issue in opportunistic networks. Recent studies show that human mobility has a high degree of temporal and spatial regularity, their movements and activity patterns which are strongly impacted by their social attributes are not aimless and chaotic. Therefore, human mobility patterns and social attributes can be exploited to optimize routing decisions in opportunistic networks.This dissertation first exploits human mobility patterns to identify important nodes for data delivery. Secondly, the non-linear time series prediction model is proposed to predict node’s future importance. Finally, this dissertation proposes two social-based data delivery algorithms for data forwarding and data dissemination in opportunistic networks. The main research contents and innovations are as follows:(1) Node centrality metrics in opportunistic network. In opportunistic networks, discovering influential nodes is one of important issues for efficient data delivery. Although traditional centrality definitions give metrics to identify the nodes with central positions based on static network model, they cannot effectively identify the influential nodes for data delivery in opportunistic networks. Then, this dissertation proposes some metrics for discovering influential nodes in opportunistic networks. This dissertation first uses the temporal evolution graph model which can accurately capture the topology dynamics of the opportunistic network over time. Based on the model, this dissertation explores human mobility patterns and social ties to extend three common centrality metrics:degree centrality, closeness centrality and betweenness centrality. The experiments with real world bluetooth data sets show that the proposed centrality metrics can well identify influential nodes in opportunistic networks.(2) Non-linear time series prediction model for node’s future importance prediction in opportunistic networks. Since the ranking of influential nodes in opportunistic networks often changes when network topology evolves over time, previous studies exploit several intuitive and simple prediction functions to predict node’s future centrality. However, the evolution of node’s centrality is not linear, high prediction errors are inevitable when the linear prediction model is used to approximate the nonlinear relation as a linear relation. Then, this dissertation proposes the non-linear time series prediction model based on Back Propagation Neural Network model which is widely used as nonlinear prediction model to predict node’s centrality in the future. This dissertation employs real world data set to evaluate the effectiveness of the proposed prediction model. Results demonstrate that the proposed prediction model is more efficient to predict node’s future centrality than linear prediction model.(3) Social-based data forwarding algorithm in opportunistic networks. Data delivery is a challenging issue in opportunistic networks because of time-varying links and intermittent connectivity. Exploring important nodes and close relations between nodes can improve the efficiency of data forwarding. Thus, this dissertation explores human contact patterns to introduce a metric to detect nodes’relations and proposed social-based forwarding centrality metric to identify influential nodes in opportunistic networks. Then, this dissertation introduces a routing algorithm based on the aforementioned metrics in which each node forwarded their messages to more central and closer relay nodes with higher probability to meet the destination. This dissertation evaluates the proposed algorithm through trace driven simulations using two real world data sets. The results demonstrate that it performs the best among the four existing algorithms.(4) Social-based publish/subscribe routing algorithm in opportunistic networks. Since exploiting humans’encounters to share messages they are interested in is one of content-based network services, publish/subscribe as a promising solution can disseminate data in interest-based scenarios. However, traditional publish/subscribe routing algorithms can’t work well in opportunistic networks due to intermittent connectivity. Thus, this dissertation exploits human mobility patterns to study publish/subscribe routing algorithm in opportunistic networks. This dissertation first proposes global entropy centrality to explore influential nodes as broker nodes for messages’publication or subscription in opportunistic networks. Since human behaviors exhibit community characteristics, this dissertation proposes the social-based distributed community detection algorithm to form friendship community for each node. Based on the above, this dissertation proposes the social-based publish/subscribe routing algorithm for data dissemination in opportunistic networks. In evaluations, this dissertation compares the algorithm with two exist publish/subscribe routing algorithms with real world traces and synthetic mobility traces. The proposed algorithm achieves the best performance.