Mobile crowd sensing: enabling technologies and applications

Nowadays, smart devices and smartphones sport a considerable number of embedded sensors which allow them to be able to sense the environment and to collect data about the context and the behavior of users. Smart devices constitute an amazing tool that can be leveraged to collect and analyse local data and, in general, to grasp a clear understanding of the complex contexts we live in. In light of this, a new fast-growing sensing paradigm has started gaining widespread adoption, leveraging the extensive presence of mobile personal devices in our lives and social participation of volunteer citizens. This new paradigm, called Mobile Crowd Sensing (MCS), has been widely adopted in distributed problem-solving applications, involving online or offline crowds. In this work, we examined the Mobile Crowd Sensing paradigm. We started analysing the process and the research progresses that led to a modern and shared definition of what MCS is and what differentiates it from other types of distributed sensing and processing initiatives (e.g., crowdsourcing, participatory sensing, etc.). Then, we surveyed a number of applications in order to pinpoint the distinguishing features of most MCS platforms and to model their common high-level process architecture. Throughout this work, we separately examined three open issues, every time trying to look at them from a fresh perspective and always proposing original solutions. We first focused on data quality. We tried to solve the common issue of having to quantify MCS data quality in numerical terms. Then we proposed a new map-matching algorithm to be applied on dense traces of MCS data, whose aim is to remove errors derived from low-quality GPS recordings. After that, we tackled the problem of providing users with substantial incentives in order to engage them in the sensing endeavour. Cooperation incentives have been in the limelight for about thirty years, but we cast the problem of finding ways to motivate users to collaborate toward a shared cause as a privacy issue. We provided a classification of most common rewarding schemes in terms of motivation supplied and anonymity of user information. We found that gamification techniques and voucher distribution are mechanisms suitable to be used in anonymous systems to engage users, thus we tested both these mechanisms. We provided and analysed a real-world design of a gamification layer built on top of an MCS application implementing strict privacy-preserving techniques. We also introduced a novel voucher-based rewarding platform, explicitly designed for MCS endeavours, that acts as a bridge between volunteers, MCS applications, and third-party stakeholders allowing the latter to be part of the rewarding paradigm without the need of participating in the crowdsensing procedure. In the last part of this work we discussed client UI approaches used in MCS instruments. The client interface is the sole point of interaction between users and platform for most crowdsensing applications. Since having a suitable number of volunteers is fundamental in MCS, joining a sensing task should not be a burden. The client UI should not constitute a barrier to platform adoption by users but, unfortunately, most of the MCS platforms used in real-world application still struggle to provide a usable and easy-to-understand UI. We presented a novel approach aiming to exploit the popularity of IM platforms, and the flexibility of the conversational interfaces they host, to empower MCS volunteers with an effective tool for contributing to the sensing task. We first studied the new phenomenon, outlining its key features, pros and cons. Then, we implemented a proof of concept in the form of a simple MCS application offering a conversational interface as client. We demonstrated the feasibility of our approach, analysed the prototype’s interface and drew conclusions on the proposed approach. In summary, we introduced and discussed many solutions to both well-known and relatively new issues in the context of MCS. Trying to keep their treatise as clear as possible, we examined each approach in isolation, providing actual implementations or at least concrete designs of our ideas. Nonetheless, discussed methods are inherently general and could be applied in many different real-world scenarios.