Research On Incentive Mechanism And Task Assignment Method In Crowdsensing Systems

Crowdsensing is a new data collection paradigm that regards distributed ordinary users and mobile devices like smartphones as basic sensing units.Compared with conventional sensor networks,crowdsensing dispenses with particular deployment and maintenance.The recruited participants usually carry the sensing equipment with better sensing,computing and storing capabilities,which allows the collaboration of implicit opportunity sensing and explicit participatory sensing to provide large-scale sensing data and computing services.This feature could be comprehensively applied in environment monitoring,public security,smart traffic and other fields,and gradually becomes an essential part of the Internet of Things system.With the popularity of portable smart devices and the development of the Internet of Things,once a large-scale crowdsensing market is formed,the design of crowdsensing systems is supposed not only to adapt to more real-time sensing databased applications but also to consider the personalized features for participants and sensing tasks such as diversity and heterogeneity.In the research of crowdsensing,considerable attention should be paid to the participating willingness and decent participants’ choice.Based on the analysis of the characteristics of the real-time crowdsensing system,this thesis accomplished systematic work on the incentive mechanism and task allocation method in the real-time crowdsensing system.The main research results and contributions are as follows:(1)An incentive mechanism based on non-cooperative games for heterogeneous multi-task scenariosIn consideration of the heterogeneous tasks led by different sensing applications and the variety of smart devices and their sensing capabilities,a market-driven unified crowdsensing framework is proposed for this heterogeneous multi-task scenario.The essence of the framework is an incentive mechanism based on non-cooperative games,supporting the platform to provide optimized sensing strategies(including task assignment and pricing guidance)for all participants(data providers and data requesters).First,based on the non-cooperative game among the crowdsensing participants,the problem is formulated as a Nash equilibrium problem.Then,a scenario decomposition method with constraints would divide complex situations into three types by complexity,which is the preparation for solution algorithm.In the end,the mechanism proposed an iterative solution to solve the stable sensing strategy so that all the participants could maximize their utility.A large number of simulations prove the effectiveness of the proposed method.(2)A task allocation method considering data freshnessAccompanied with the abundance of the types of sensing tasks with data timeliness sensitivity,a task allocation method is proposed with consideration of data freshness to improve the sensing data quality and the willingness of participants.Based on the incentive mechanism described in the first study,age of information(Ao I)is initially introduced to measure the freshness of data in crowdsensing,hence a data quality model is established.Two situations are distinguished according to the sensitivity.For the Ao Isensitive case,the sensing strategy would plan the sensing time for sensing tasks and the executing sequence of the sensing tasks according to data value variation,namely,it includes the deterioration of data timeliness.A coevolutionary algorithm is designed to solve the sensing strategy for each provider to achieve Nash equilibrium.Meanwhile,for the Ao I-insensitive case,an improved relaxation algorithm is proposed to provide a stable strategy for the data provider.Numerical findings reveal that the developed algorithm can effectively identify stable sensing strategies in a Nash equilibrium.(3)A sensing coverage quality enhancement method based on mobile scheduling in target areaThe data collection process from the opportunity sensing method cannot guarantee a timely response to the sensing demands randomly happening in the target areas.Especially,for the time-sensitive sensing tasks,waiting for rational movements passing target areas would lead to delay with significant influence.Consequently,the article explores the possibility of increasing the coverage of target areas in time-sensitive scenes through participatory sensing methods.Firstly,a framework is proposed to solve the problem of sensing task allocation with dynamic sensing coverage quality.Then a coverage quality model based on area sensing coverage density and an incentive model for participants is established to assist the solution.Since determining the optimal allocation solution is an NP-hard problem,here are two approximate algorithms to solve it.The experimental results show that the performance of the two algorithms is significantly superior to that of the baseline algorithm.In conclusion,this thesis proposes a crowdsensing system framework with a reasonable incentive mechanism and an effective algorithm.The corresponding theoretical analysis and extensive experiments verify that the designed incentive mechanisms have reached the expected goal effect.The result of the thesis provides an essential theoretical foundation and technical guidance for the deployment and application of a real-time crowdsensing system.