Green CrowdSensing with Comprehensive Reputation Awareness and Predictive Device-Application Matching Using a New Real-Life Dataset

Mobile CrowdSensing (MCS) has emerged as a valuable framework for large scale mapping of phenomena of interest, thanks to the ever-growing advances and pervasiveness of sensor-rich mobile devices. Meanwhile, towards building a robust MCS system, several challenges have yet to be overcome. This includes security challenges, privacy concerns, data integrity, in addition to others. In this research, we are concerned with the power consumption of sensing campaigns, from the perspectives of service demanders and participants. We propose an opportunistic and predictive crowdsensing management framework that realizes Green Mobile CrowdSensing (GMCS) campaigns through energy-aware participant-task matching. This is achieved using two new techniques. First, we propose Green Auctioning, an auction management technique which adopts a new objective function that guarantees the selection of the devices which consume the least energy. The proposed objective function features a hard reputation term which depends on the device’s energy consumption, in addition to the soft reputation term previously proposed in the literature. Second, we propose Predictive Auctioning, an auction management technique that adopts machine learning models to empower the platform to predict users’ ability to complete the task at hand, using the user’s battery level and internet connection status, and the task’s duration. Towards this goal, we have constructed a new dataset–the MAGGIE dataset–by monitoring the energy consumption of over 100 users, over a duration of four months, using a mobile application specially developed for this purpose. To the best of our knowledge, this is the first research that addresses energy-aware auctioning by constructing a crowd-sensed dataset specially built for this purpose. We present promising results for the attained energy awareness without compromising other performance aspects including data trustworthiness and the clearance rate of sensing auctions.