Research On Incentive Mechanism Of Mobile Crowdsensing

The development of communication technology and sensor technology has enabled more and more and more powerful computing,sensing,storage and communication capabilities of smart mobile devices.With the explosive popularization of these smart mobile devices,a new sensing paradigm,namely Mobile Crowdsensing(MCS),has emerged.MCS combines the idea of crowdsourcing with mobile sensing.Benefiting from the continuous movement and extensive participation of a large number of intelligent mobile devices,MCS can provide large-scale complex sensing.Compared to traditional sensor networks,MCS can achieve lower deployment cost,faster system construction,easier maintenance,more scalability,high coverage and high mobility,and has a wide range of applications.The efficient implementation of MCS relies on the cooperation between a large numbers of participants.However,sensing activities will consume certain resources of participants and at the same time there is a risk of privacy leakage,which leads to lower willingness for participants to deal with sensing tasks.Therefore,it is essential to design efficient and reasonable incentive mechanisms to support the implemention of MCS system.Due to the constant movement of participants and the fact that the arrival of sensing requests are highly stochastic and difficult to predict,it is extremely challenging to design a reasonable and efficient incentive mechanism to maximize the system utility.This paper introduces Lyapunov’s optimization theory as a tool to carry out researches on the design of incentive mechanisms under complete information or incomplete information.The main work and contribution of this paper include five aspects,and they are shown in the following contents,respectively.(1)The MCS incentive mechanism framework is designed.First,the importance and necessity of MCS incentive mechanism framework is analyzed based on the study of the existing MCS framework.Due to the lack of the MCS incentive mechanism framework,the incentive mechanism framework is designed from three aspects: basic definitions,components,and selection framework.The definition and description method of the incentive mechanism are given,including not only the traditional equality or inequality constraints,but also property constraints such as individual rationality,incentive compatibility,computational effectiveness,stability,and system effectiveness.Next,the components of the incentive mechanism are detailly analyzed,including six aspects: incentive objective,interaction mode,incentive means,constraint,participant selection,and incentive allocation,to help researchers understand the MCS incentive mechanism systematically.Finally,an incentive mechanism selection framework P4 is designed to guide the service platform to find suitable incentive mechanisms under different conditions.(2)A Cooperative Incentive Mechanism under Complete Information(CCIIM)based on current information and Greedy Predictive Cooperative Incentive Mechanism under Complete Information(GP-CCIIM)based on greedy predictive information are proposed.First,the incentive mechanism design problem is transformed into a problem that minimizes an upper bound of drift-minus-utility to reduce the complexity of solving the problem,and then CCIIM is designed.CCIIM can make decisions only based on the currently available information of the system.Next,the prediction information is incorporated into CCIIM,and the Predictive Cooperative Incentive Mechanism under Complete Information(P-CCIIM)is proposed.In order to reduce the computational complexity of P-CCIIM,this paper further proposes the GP-CCIIM incentive mechanism.Theoretical analysis shows that under any control parameter V > 0,the difference between their average utility achieved by CCIIM and P-CCIIM and the optimal utility can be bounded within O(1/V),while the average backlog can be bounded within O(V).Extensive numerical results based on synthetic data and real trace data show that CCIIM outperforms Greedy incentive mechanism and Random incentive mechanism,and GP-CCIIM improves the utility-backlog tradeoff of CCIIM.(3)A Non-cooperative Incentive Mechanism under Complete Information(NCIIM)is proposed.Based on the research of CCIIM,the competition between the service platform and the participants is taken into consideration,and the participants are involved in the design process of the incentive mechanism.First,the interaction procedure between the service platform and the participants is modeled as a two-stage non-cooperative Stackelberg game.The service platform and the participants make their own decisions to maximize their own utility.On this basis,the incentive mechanism NCIIM was designed by using backward induction.The incentive mechanism can implement only based on the current available information.Meanwhile,both theoretical analysis and simulations show the equilibrium between the platform and participants,and none of the stakeholders,including the service platform and participants,can improve his utility by unilaterally changing its current strategy.Finally,it is found that the competition between the service platform and the participants will cause a certain utility loss of the system,and this loss can become very small under certain conditions.(4)An Incentive-Compatible Incentive Mechanism under Incomplete Information(ICIIIM)is proposed.First,the interaction procedure is modeled as an auction.In the auction,the participants bids for reward and the service platform assignes the reward based on the bid.Then this paper introduces the customized Myerson theorem to ensure that the designed incentive mechanism satisfies the incentive compatibility,that is,when the participant’s bid is the true valuation of the cost,it achieves maximum utility.On this basis,the incentive mechanism ICIIIM is designed.ICIIIM can simultaneously achieve the properties of individual rationality,incentive compatibility,computational effectiveness,stability,system effectiveness,etc.Extensive experiments show that ICIIIM achieves higher system utility and platform utility than the classical incentive mechanism IMCU.(5)An Incentive Mechanism under Incomplete Information and Participatory Constraint(PCIIIM)is proposed.Based on the research of ICIIIM,the participant’s participatory constraint is taken into account,that is,the probability of winning the auction for each participant should reach a certain level.First,the participation constraint is transformed to the stability of virtual queue with a regulator to represent the system’s satisfaction level.On this basis,PCIIIM is designed.PCIIIM can simultaneously aschieve the participation constraint as well as property constraints such as individual rationality,incentive compatibility,computational effectiveness,stability,and system effectiveness.Extensive experiments show that,when considering the participation constraint,PCIIIM is superior to ICIIIM and IMCU in terms of system utility,platform utility,and the number of active participants.In summary,this paper explores several aspects of incentive mechanism design in MCS and provides theoretical evidences and supports for the efficient implementation of MCS.These work has certain academic and practical significance for further research and applicaton in MCS.