Crowdsourced Network Resource Sharing: Optimization and Incentives

摘要

The proliferation of smart mobile devices has witnessed the dilemma of the coexistence of challenges and opportunities in mobile data networks. The fast growing mobile data traffic due to the bandwidth-hungry mobile devices has made the mobile networks increasingly congested. Meanwhile, the rich computing resources and the diverse sensors embedded in the mobile devices can be efficiently leveraged for large-scale mobile sensing data acquisitions. In this thesis, we address the dilemma by proposing a unified crowdsourced network resource sharing framework, and address the economic incentives issues for enabling such a framework.;In the first part of the thesis, we consider the profit maximizing crowdsourced spectrum sharing. The crowdsourced spectrum sharing is a promising way of addressing the spectrum scarcity problem, by crowdsourcing to the spectrum legacy owners for the available spectrum resources. In particular, we study how to maximize a spectrum database operator's expected profit in sharing the crowdsourced spectrum to secondary users with stochastic arrivals and heterogeneous spectrum requirements. By jointly pricing and scheduling the spectrum resources crowdsourced by the spectrum legacy owners, we show that the optimal dynamic pricing can improve the spectrum database operator's profit by more than 30% over the optimal static pricing, when secondary users' demands are highly elastic. Our results show that such a profit maximizing crowdsourced spectrum sharing can significantly increase the sharing network performance.;In the second part of the thesis, we consider the cost-efficient crowdsourced network resource sharing. The crowdsourced resource sharing of mobile devices is a promising way of performing cost-efficient large-scale mobile sensing, by crowdsourcing to the mobile devices for the embedded sensors and computing resources. In particular, first, we present a scalable mobile crowd sensing system via peer-to-peer data sharing, to alleviate the high operational cost and the poor scalability of the traditional server-client mobile sensing system. In the proposed sensing system, the sensing data is saved and processed in user devices locally, and is shared among users in a distributed manner. To incentivize the data sharing, we propose a data sharing market with revenue sharing to the sensing users. The analysis shows that a significant system efficiency can be achieved by properly choosing the revenue sharing strategies. Second, we propose a cost-efficient three-layer data-centric mobile crowd sensing model by introducing a new data layer between sensing tasks and mobile users. Such a data-centric mobile crowd sensing model enables different sensing tasks to reuse the same data items, hence can effectively leverage the task similarity. We first analyze the centralized optimization problem and show that the performance gain due to data reuse can be quite significant. Then we consider the two-sided information asymmetry of selfish task owners and users, and propose a decentralized market mechanism for achieving the centralized optimality. Our results show that such a cost-efficient crowdsourced sharing can significantly decrease the operational cost and achieve more efficient resource utilizations.