On The Platform Profit Maximizing In Crowdsensing System

With fast development of smartphone and communication network,smartphone users who are able to acquire all kinds of data by advanced smartphone sensors.Compared to the traditional collecting sensing data paradigm,this new approach enjoys several great advantages,such as lower cost,more flexible and gigantic data volume.These characteristics provide a new way to deal with environment problems,traffic conditions,public health and security problems and so on.This is where the concept crowdsensing derives.A typical crowdsensing system involves three roles,data requesters,smartphone users and a platform which organizes transactions.As smartphone users will endure costs while providing sensing data,they are reluctant to provide their sensing data.Therefore,a lot of research works focus on the incentive mechanism that stimulates smartphone users while guarantee the social welfare or the profit of data requesters.However,few research works have considered the profit maximization of the platform.In this thesis,we study the auction incentive mechanism and propose two models of platform profit in crowdsensing system: a one-stage platform profit model and a multi-stage platform profit model.In the one-stage model,we formulate the model,solve it,and propose appropriate algorithms to maximize its profit.For the multi-stage model,we focus on the longterm profit of the platform.According to the solution of the model and simulations,we find that the platform will sacrifice benefit from one transaction in order to stimulate the participation of crowdsensing system.To be more specific,for the one-stage platform profit model,the participants are only allowed to have one-time transaction.We start from analyzing the basic constraints of a valid auction incentive mechanism,modeling the interaction between three roles as a Bayesian game.After mathematical deduction,we obtain the expression of platform profit.We then transform the platform profit maximizing problem to an optimization problem and prove that it is an assignment problem.At last,we propose two solutions,one exact algorithm and one approximation algorithm.For the multi-stage platform profit model,we firstly introduce the concept of time slot.During each time slot,data requester and smartphones will conduct transactions via platform.The participants will face a decision problem whether to stay on the platform for the next period.We use the dynamic programming approach to analyze the utility function of participants and propose optimal strategies for them.We obtain by consequence the probability of staying on the platform of all participants given a time slot.Based on that,we are able to build the multi-stage platform profit model.The platform can adjust payment from data requesters and payment to smartphone users so as to achieve the maximum profit.At last,we select suitable datasets to fit the crowdsensing task value and cost,which are random variables.Consequently,we can run simulations for our two platform profit models.According to the results,the effectiveness of our solutions is confirmed.