CrowdBind: Fairness Enhanced Late Binding Task Scheduling in Mobile Crowdsensing

Mobile crowdsensing (MCS) is an efficient method to collect sensing data from a large number of mobile devices. Traditionally, low task coverage and high energy consumption on mobile devices are two of the main challenges in MCS and they are extensively studied in the literature. In this work, we discuss a third factor, scheduling fairness, which is correlated with the other two factors and has a significant impact on the success of MCS. We propose a new framework, called CROWDBIND, that takes advantage of the latebinding characteristic of crowdsensing tasks in addition to incorporating a trajectory-based mobility prediction model to schedule tasks. We conducted a survey with 96 participants to learn about how users react to varying levels of fairness in MCS applications. We designed and implemented a full-stack MCS system including a scheduling server and an Android client. We evaluate our system by conducting an IRB approved user study of 50 people in our college town for one month as well as running a simulation using Gowalla dataset of 90K users. CROWDBIND is proved to be effective in a large population and the results show that CROWDBIND achieves the highest scheduling fairness compared to prior works (Periodic sensing, PCS, Sense-Aid, and CrowdRecruiter), improves the average per-device energy efficiency from 18.3% to 91.4%, and improves the task coverage from 9.7% to 52.1%.