Research And Implementation Of Quality-aware Incentive Methods In Crowdsensing Applications

CS applications recruit normal people as participator to collect sensing data,which are used to complete large-scale and complex social sensing tasks.Considering the cost of task execution and data uploading,few mobile users are willing to participate in a CS system voluntarily.However,for a sensing platform,user participation is crucial.Therefore,designing an effective incentive method to compensate users is necessary.The sensing quality of every user varies significantly due to various factors.Therefore,an incentive method should not only guarantee the payment to users,but also try to obtain high quality data with little expense.Though service quality of each individual mobile user is considered in traditional incentive methods,marginal effect in data collection is ignored.When it comes to online user recruitment,while the dynamic participation of users is considered in previous works,users' sensing qualities are ignored.Aiming at solving the incentive problem in CS,the paper used game theory to propose two quality-aware incentive methods which can select users based on sensing quality.These two methods both have individual rationality,truthfulness and computational efficiency.Following research works has been done in this paper:1.Offline incentive method.Offline means a sensing platform choose users after they submit their bids.In our method,we consider one user can participate multiple tasks and the platform will calculate the sensing quality of the whole crowd in winner selection to handle with marginal effect of sensing data.We improved the traditional one-parameter truthful mechanism to apply two-parameter CS.On that basis,we presented QIM,which can produce close-to-optimal solutions but run faster.2.Online incentive method.Online method can handle dynamic participation.We hope the sensing platform can make user selection decision without requiring previous knowledge of users.Therefore,we used the expected quality to design a monotone user selection rule and used transformation proposed by Babioff to transform it into a truthful mechanism QOIM.3.We conducted simulation using the mobility dataset of San Francisco taxies to compare QIM and other two methods.For QOIM,we used real trajectory sets from the Dartmouth College mobility traces to do simulation with our method and another method.Finally,we compared QIM and QOIM and did some analyzation about them.