RECrowd: a reliable participant selection framework with truthful willingness in mobile crowdsensing

Abstract

In crowdsensing systems, participant selection as one of main problems attracts a lot of attention. Most studies focus on how to select reliable participants for task allocation to assure task quality and minimise incentive cost. Conventional methods are mainly based on historical reputation, which is from the statistical result of participant behaviours during a past period. However, these methods could cause unreliable assignment that reduces the task quality, since historical reputation cannot exactly reflect the current state of participants. In this paper, we advocate RECrowd, a reliable participant selection framework that considers both historical records and current truthful willingness. In RECrowd, we formulate an optimisation problem with the objective of minimising incentive cost while ensuring task quality, and design a two-stage online greedy algorithm with the pre-assignment step. During the participant selection, we also consider participants' position privacy. Experiment results with real datasets demonstrate our algorithm outperforms other methods.