Auction Based Task Assignment Mechanism Design For Crowdsourcing Systems

In recent years,as a new paradigm for data collection and problem solving,crowdsourc-ing has been widely applied to various fields.However,these crowdsourcing applications,harnessing the intelligence of a large crowd of human workers,have been severely affected by the limited number of task participants and the low quality of the data services provided.To tackle the problem,auction is a common way used in crowdsourcing task assignment to stimulate participants to provide high quality answers.This thesis focuses on auction based task assignment mechanism to incentive partici-pants with monetary rewards.By analyzing crowdsourcing systems in different scenarios,we design a truthful and efficient task assignment mechanism,which mainly includes two parts:allocation algorithm and pricing(payment)algorithm.The contributions of this thesis are summarized as follows.(1)Considering budget constraint of the requester and the heterogeneity of task,by con-structing the reverse combination auction model,we proposed TCAM mechanism to maximize the requester's profit on the basis of sub-modular function.Through theoret-ical analysis and extensive simulation,it proves that the proposed mechanism satisfies individual rationality,truthfulness and budget feasible,motivates participants to provide high-quality data,and maximizes the profit of requester.(2)Considering that there are multiple task requesters in crowdsourcing system,we de-signed a truthful and efficient multi-unit task assignment mechanism by constructing adouble auction via low bid priority matching and uniform pricing principle.Through theoretical analysis and experimental verification,it is proved that the proposed mecha-nism satisfies three economic properties:truthfulness,individual rationality and budget balance.(3)Considering the scene that workers randomly arrive at crowdsourcing platform online,by constructing an online auction model and adopting a two-stage selecting process to design an online allocation mechanism for homogeneous tasks for the first time.Through theoretical analysis and simulation results,the mechanism can stimulate high-quality workers to participate and improve the completion quality of task under the basis of satisfying the three economic properties:individual rationality,truthfulness and bud-get feasible.The above crowdsourcing task allocation models combine auction to design incentive mechanisms,applicable to the single-requester,multi-requester and online scenario of the crowdsourcing system respectively.After theoretical analysis and simulation results,the above three models have achieved the effect of motivating participants to improve the com-pletion quality of crowdsourcing tasks.