ProSC+: Profit-Driven Online Participant Selection in Compressive Mobile Crowdsensing

Abstract

A mobile crowdsensing (MCS) platform motivates to employ participants from the crowd to complete sensing tasks. A crucial problem is to maximize the profit of the platform, i.e., the charge of a sensing task minus the payments to participants that execute the task. Recently, the appearance of data reconstruction method makes it possible to improve the platform's profit with a limited amount of sensing results in Compressive MCS (CMCS). However, It is of great challenge to the maximal profit for the CMCS platform, since it is hard to predict the reconstruction quality due to the dynamic features and mobility of participants. In response to such challenges, we propose two profit-driven online participant selection mechanisms for the given task model and participant model. In ProSC, the sub-profit in each slot is maximized during the sensing period of a task, by combing a statistical-based quality prediction method and a repetitive cross-validation algorithm. In ProSC+, we jointly optimize the number of required participants and their spatial distribution to further improve the converging property. Finally, we conduct comprehensive evaluations, the results indicate the effectiveness and efficiency of our mechanisms.