Summer projects
Picking Digital Pockets project
Project Overview
Mobile devices of people form a new and unexplored data reserve. In an urban area, terabytes of data are literally walking around in the pockets of people, but no means exists to take ad- vantage of the availability of this data in the environment. This is in stark contrast to other data reserves for which (central) indexing and mining techniques are well established, the most prominent global resource being the union of all (publicly) accessible web pages and their contents. Mining mechanisms are also progressing for other resources including online social networks such as FaceBook or LinkedIn (access to which is limited) and user-generated content databases (e.g., YouTube, Flickr). Finally, closed databases, e.g., of fixed and mobile operators as well as third parties exist containing information about user mobility and communication behavior. All these have in common that they are - for an authorized party - easily accessible so that the evaluation can focus on interpreting the data, the main difficulties including the management of vast amounts of data and inference of interesting properties.
The situation is fundamentally different for human users' mobile devices (their digital pockets): these are not (well: should not be) easily accessible from the outside, they are not always online (for wireless coverage, cost, or energy reasons), and their data is not necessarily linked in any structured manner as web pages are (but rather spread across different storage locations, e.g., per application) and thus also not easily subdivided in private and public parts. At best, different applications (possibly running on dedicated computers) may have partial and likely outdated views of the mobile's content (e.g., iTunes for music and videos, iPhoto for images, dropbox for sharable files, a calendar for appointments, etc.). This application knowledge is usually not explicitly exported either, for good reasons. Therefore, content on mobile devices is virtually unexplored from a data indexing, mining, and sharing perspective.
In the PDP project, we plan to explore mechanisms for searching and mining these kinds of data reserves - in short: picking digital pockets - and develop a fundamental understanding about the nature of such mobile data reserves and how their nature affects applications de- signed to exploit them. We recognize the special character of mobile data reserves for the aforementioned reasons and devise a fully decentralized system in which a mobile node's geographical environment becomes the region of interest for searching, mining, and sharing. We assume that general searches would usually be carried out via the Internet (while connected), whereas content of local relevance would be available in the (immediate) surroundings and thus naturally be obtained from neighboring or close-by mobile nodes. The latter interaction would particularly take place using ad-hoc node-to-node communication capabilities for at least two reasons: (1) Not using a central element for sharing avoids both censor- ship and data gathering so that important properties such as anonymity and privacy can be retained. (2) Ad-hoc communication allows offloading data from the wireless access links, thereby aiding in conserving scarce link capacity and supporting operation even if there is no (affordable) wireless network infrastructure available.
Available Topics
We have multiple possible topics around the PDP project and they can be tailored to the skills of the student. Some examples include:
- Implementation of a mobile client for PDP
- Simulating large PDP environments
- Adaptive search in mobile opportunistic networks
- Routing in cognitive radio enabled mobile opportunistic networks
MOSES: Mobile Opportunistic Services for Experience Sharing
The project pursues the vision of augmented reality for visitors in co-located social communities for visitors of entertainment venues such as the Expo 2015 fair. The project will extend an existing mobile communication platform that readily supports mobile opportunistic networking and service provisioning and develop applications that use those features to realize instant experience sharing. The aim is to produce designs and prototypes right from the beginning to ensure that multiple applications will be deployed, trialled, evolved, and evaluated during the project.
The outcome enables visitors to experience more than their immediate surroundings instantaneously and assist them in dynamically (re)organizing their visit due to comprehensive ways to interact with each other. They can communicate within their group, but also with people near them to coordinate, e.g., to play instant games and to participate in others' experiences. Visitors be able to form co-located social communities to share the moment, issue hints about cool things they have seen, and enrich each other's experiences with collaborative games and shared media. Remote parties such as family and friends at home can also join in some elements of play.
In order to realize this vision, it is necessary to develop support for applications that enable users to efficiently share potentially large-size multimedia content between them, and venue operators to collect detailed data about the status of the venue during visits. It is becoming obvious that existing cellular operator infrastructure in many cases is not able to cope with the explosive growth in mobile data traffic. WiFi offloading helps and is further enhanced by new direct peer connection technologies which are just becoming viable but for which best usage patterns are still poorly understood. Not only direct peer connections (supported by opportunistic or delay-tolerant communication paradigms) can alleviate infrastructure congestion. They can enable brand new types of services for multimedia applications, by exploiting embedded knowledge about physical proximity of the users and common location (to be possibly correlated with users' profile information to better understand the dynamics of social interactions).
Available Topics
Topics will be around implementation and testing of the MOSES platform and require strong coding skills for small and mobile devices.
Miscellaneous Projects
We also have other topics that are not directly related to actual projects but around which summer job topics can be defined.
Available topics revolve around implementation of different kinds of opportunistic sensing and sharing systems, such as:
- Poor Man's Surveillance Camera (using a Canon camera and a development kit with associated computers)
- Extending our existing Raspberry Pi testbed by developing new services
- Also see recent open MS thesis topics under my name