PhD Oral Exam - Gang Li, Electrical and Computer Engineering

Auction-based Efficient Online Incentive Mechanism Designs in Wireless Networks


Recently, wide use of mobile devices and applications, such as YouTube and Twitter, has facilitated every aspect of our daily lives. Meanwhile, it has also posed great challenges to enable resource-demanding users to successfully access networks. Thus, in order to enlarge network capacity and fully make use of vacant resources, new communication architectures emerge, such as D2D communications, edge computing, and crowdsourcing, all of which ask for involvement of end mobile users in assisting transmission, computation, or network management. However, end mobile users are not always willing to actively provide such sharing services if no reimbursements are provided as they need to consume their own computation and communication resources. Besides, since mobile users are not always stationary, they can opt-in and opt-out the network for their own convenience. Thus, an important practical characteristic of wireless networks, i.e., the mobility of mobile users cannot be ignored, which means that the demands of mobile users span over a period of time. As one of promising solutions, the online incentive mechanism design has been introduced in wireless networks in order to motivate the participation of more mobile users under a dynamic environment. In this thesis, with the analyses of each stakeholder's economic payoffs in wireless networks, the auction-based online incentive mechanisms are proposed to achieve resource allocations, participant selections, and payment determinations in two wireless networks, i.e., Crowdsensing and mobile edge computing. In particular, i) an online incentive mechanism is designed to guarantee Quality of Information of each arriving task in mobile crowdsensing networks, followed by an enhanced online strategy which could further improves the competitive ratio; ii) an online incentive mechanism jointly considering communication and computation resource allocations in collaborative edge computing networks is proposed based on the primal-dual theory; iii) to deal with the nonlinear issue in edge computing networks, an nonlinear online incentive mechanism under energy budget constraints of mobile users is designed based on the Maximal-in-Distributional Range framework; and iv) inspired by the recent development of deep learning techniques, a deep incentive mechanism with the budget balance of each mobile user is proposed to maximize the net revenue of service providers by leveraging the multi-task machine learning model. Both theoretical analyses and numerical results demonstrate the effectiveness of the designed mechanisms.