Crowd-sensing: Human-powered Sensing in Opportunistic Mobile Social Networks

Project description

In this project, we focus on a framework to characterize and leverage people movement to improve sensing tasks in opportunistic mobile social network. As the first part of the project, motivated by the lack of sensing devices in crowd event, we attempt to solve the problem of recovering the wireless contacts between non-sensing participants. Particularly, we propose a context-aware approach that incorporate contextual information of the study (e.g., length of sensing interval, density of the crowd) into a classification model, together with other pairwise features between two participants, to classify if there is a missing contact between them or not. The results show that the context information plays an important role in identifying potential missing wireless contacts.

For the second part of the project, we aim to improve the way sensing tasks are deployed in crowd events. Traditionally, the sensing devices are usually distributed and configured in an ad-hoc manner. For example, the number of devices is fixed, people register for the devices on a volunteering basis, and the sensing interval is static. We, on the other hand, consider all those assumptions as costs or constraints and model the crowd-sensing problem as an optimization task. We propose a two-stage approximation algorithm with bootstrapping and contextual feedback adjustments to solve the optimization problem.