Research On Data Collection And Incentive Mechanisms In Mobile Crowd Sensing Networks

Mobile Crowd Sensing (MCS) networks treat mobile devices of ordinary users as basic sensing units, and leverage them to distribute sensing tasks and collecting sensing data by consciously or unconsciously collaborating with each other through the mobile Internet for completing large-scale and com-plex social sensing tasks. They not only provide a novel sensing paradigm for the Internet of Things, but also bring a series of new challenges, which become a hot research topic in recent years. Compared with traditional Wireless Sen-sor Networks (WSNs), one of the most important features of MCS networks is the human-centric paradigm, namely that humans will participate in the whole sensing process as both of "consumers" and "producers" of sensing data. Although currently there are variety of MCS application systems, it still lacks basic models of evaluating the data collection quality, efficient data collection methods, and reasonable incentive mechanisms for attracting user participa-tion. In traditional WSNs, the data collection problem has been well investi-gated, and there are many existing methods, but it is difficult to apply these methods to MCS networks directly, and the user incentives are not considered.Thus, this thesis focuses on the data collection and the related participation incentive problems in MCS networks, and proposes a serious of novel models and methods from three perspectives:how to measure and analyze the data collection quality, how to design efficient data collection methods, and how to provide incentives to users for participating in data collection. The main contributions of this thesis are as follows:(1) Measurement model and analysis method of coverage quality. As an important performance metric of measuring the data collection quality, the coverage in MCS networks is different from that in traditional WSNs, which is closely related to the opportunistic nature of human mobility. Considering the time-varying factor of sensing coverage in MCS networks, we propose the inter-cover-time as the measurement metric. Based on the datasets of taxi mo-bility traces in Beijing and Shanghai, we analyze the distribution of inter-cover-times, and derive the relationship between the coverage ratio and the number of nodes. The proposed measurement model and analysis method of coverage quality can provide good theoretical guidelines on network planning.(2) Cooperative opportunistic sensing based on the spatio-temporal corre-lation. The energy saving problem is an important challenge faced by the data collection in MCS networks, so we design a cooperative opportunistic sensing framework. Firstly, we propose an offline node selection mechanism to select the minimum number of nodes to achieve coverage requirements, based on the history trajectories of the given set of nodes. Secondly, we propose an online adaptive sampling mechanism to help each selected node to decide whether to perform the sampling task at some time adaptively. Extensive simulation re-sults verify that our proposed mechanisms can guarantee the data collection quality, and reduce the sensing energy consumption.(3) Cooperative opportunistic transmission by leveraging data fusion. Ex-isting opportunistic forwarding schemes only focus on the sharing and dissem-ination of sensing data interested by individuals, but do not consider the spatio-temporal correlation of sensing data. We design a cooperative opportunistic transmission framework to improve the network transmission performance by integrating the opportunistic forwarding schemes and data fusion. Based on this framework, we propose two cooperative forwarding schemes:Epidemic Routing with Fusion (ERF) and Binary Spray-and-Wait with Fusion (BSWF), derive the dissemination law of correlated packets, and design new data for-warding rules. Extensive simulation results verify that our proposed mecha-nisms can guarantee the data collection quality, and reduce the transmission energy consumption. (4) Online incentive mechanisms for attracting user participation. Most of existing incentive mechanisms are offline, which assume that all of interested users submit their profiles to the crowdsourcer in advance. In practice, howev-er, users always arrive one by one online in a random order. We study online incentive mechanisms to enable the crowdsourcer to select a subset of users before a specified deadline, so that the value of services provided by selected users is maximized under a budget constraint. By considering the character-istics of different users, we investigate the case where the value function of selected users is non-negative monotone submodular, which covers many re-alistic scenarios. Based on the online auction model, we propose two online incentive mechanisms, OMZ and OMG, in the zero arrival-departure interval model and the general interval model respectively. We prove that the two mech-anisms can satisfy the computational efficiency, individual rationality, budget feasibility, truthfulness, consumer sovereignty, and constant competitiveness by theoretical analysis and experiments.In summary, this thesis not only proposes a series of models and meth-ods on data collection and incentive mechanisms, but also provides theoretical analysis and extensive experiments to verify the effectiveness of them, pro-viding important theoretical and technical support for the wide applications of MCS networks.