Research On Message Forwarding In Social Opportunistic Networks Based On Social Infomation

Social Opportunistic Networks (SONs) are a sort of opportunistic networks, which are composed by many nodes, e.g., hand-held mobile terminals. SONs inherit the typical characteristics of opportunistic network, e.g., frequent contacts, high transmission latency, intermittent network connecting, limited computation ability and buffer memory. Due to these features, messages usually are forwarded among the nodes in Store-Carry-Forward approach. Unlike normal opportunistic networks, the willing of people determines the mobility of nodes in SONs, which reflects sociality of nodes in mobility and contact. Social analysis is the main approach to study the nodes' behavior, while most of existing researches do qualitative analysis on the impact of social characteristics on nodes' behavior rather than quantitative analysis, which cause deviations and defects in forwarding algorithm. In addition, the majority of social characteristic based forwarding mechanisms neglect some issues, such as network congestion caused by centrality-based algorithms, buffer overflow and uneven consumption of resources, and performance degradation of the similarity based forwarding algorithm caused by unsocial nodes.This thesis is to study the SONs that made up with a large amount of hand-held devices using short-range communication, and investigate the performance of data transmission service in situation of poor network connectivity, or expensive cost of network deployment. This thesis focus on three aspects, consisting of social behavior model, message forwarding algorithm, and TTL and buffer replacement mechanism. The concrete research content is as follow:1) Based on the study on social characteristics and trace data analysis, we first verify that common social characteristics exert positive influence on nodes' contact. Then the social-characteristic-influenced contact model is proposed, compared with with random contact model, we propose a method to quantitatively compute the effect of social characteristics to contact.2) Based on the effects of social characteristics on nodes' contact, the weighted Characteristic based Forwarding (COF) algorithm is proposed, in which the messages are delivered to the nodes that are more similar to the destination node. To avoid the negative effect of unsocial nodes, we propose OCOF algorithm, in which we evaluate the centrality of nodes with message relay frequency and contact set, and the unsocial nodes are circumvented when forwarding a message. At last, we introduce a coordination factor to adjust the proportion of similarity and centrality in our algorithm to adapt to more scenes.3) According to the average delivery latency of different messages, the weighted characteristic influenced TTL allocation strategy is proposed. Different messages are allocated corresponding TTL to make rapid elimination of redundant copies. Besides, to reduce the negative impact of buffer overflow, we propose Characteristic-based Maximal Value (CMV) algorithm to make buffer replacement, in which messages with less replicas and lower expectation will be discarded first.4) Evaluate the proposed forwarding algorithm and buffer replacement mechanism with simulation experiment, the validity is verified. Compared with classic algorithms in opportunistic networks, the proposed algorithm shows advantage on performance.