Forgive But Don't Forget: On Reliable Multi-Task Allocation in Mobile CrowdSensing Platforms

In Mobile Crowd Sensing (MCS) platforms, users are typically human participants who willingly take time out of their daily schedules to complete sensing tasks. Albeit the unreliable nature of human's behavior, existing task allocation mechanisms proposed within MCS platforms typically assume that participants will accept the tasks allocated to them and complete them successfully, which in turn affects the realized quality of task completion. In this paper, we define a novel participation reliability metric, which forgives erratic misbehavior but doesn't forget if it's repeated. Moreover, to incentivize participants to be more reliable, we integrate the defined reliability metric into an online multi-task allocation mechanism, associated with a rational payment model. Finally, we theoretically analyze the proposed components and evaluate their performance on synthesized mobility traces.