The main goal of this project is to define and develop a common cloud computing framework that can be used to stimulate the design and development of the next generation of mobile applications. The project will investigate and identify the common set of new cloud services and components needed to support the growth of next-generation mobile applications. The project will develop a framework that integrates these common services, adapts them to the cloud, and interfaces easily with developers to make it convenient to build next-generation mobile cloud computing applications. This project will further explore what enhancements are needed to existing cloud platforms/infrastructures to accommodate this new mobile cloud computing framework.
This project has significant broader impacts. The outcomes of this project will be released as open source to spur the advancement of research and development in next-generation mobile cloud applications and systems. The research results of this project will be incorporated into relevant undergraduate and graduate courses. Undergraduate students will be employed in research through REU supplements and the SMART multicultural program at the University of Colorado at Boulder, and a new computation unit will be created and offered to middle and high school students.
In order to identify what common infrastructural features or services are needed to support next-generation context-aware mobile cloud applications, this research project investigated mobile cloud context in several areas. The next four paragraphs summarize the intellectual merit and findings achieved in this research, followed by a summary of the key outcomes. First, this research showed the feasibility of fusing together mobile context obtained from location, audio, and accelerometer sensors on smartphones so that unusual events can be accurately and efficiently detected using a hybrid algorithm that combines unsupervised classification of location-based audio and location-based activity with supervised joint fusion classification. Second, this project advanced the research community's understanding of context, namely normal user behavior and misbehavior, within mobile video chat applications, finding that (a) most random mobile video chat sessions were short as users sought interesting people to talk to, (b) normal users are highly correlated with using the front camera and showing their faces, whereas misbehaving users tend to hide their faces – which suggests the exploration of camera position and face detection for distinguishing normal users from misbehaving ones, (c) users with a large enough fraction of sustained sessions are disproportionately female, but surprisingly females were just as likely to misbehave as males, and (d) mobility introduces more diversity than Web online content, while groups typically imply normal behavior. In addition, the research achieved a greater understanding of how to scale cloud-based detection of misbehaving users in online random video chat. Third, the research initiated a novel study of mobile group contextual behavior to support the development of novel recommendation algorithms for mobile groups of users, investigating what factors such as mobility, separation, and social relationship would affect a group's decision to dine at a particular restaurant, and thereby influence the group recommendation algorithm. Preliminary results were obtained and are being prepared for publication. Fourth, in addition to improving understanding of mobile cloud context, the project also made progress towards developing a software infrastructure to support mobile cloud applications, developing both an architecture and an initial software prototype. Summarizing some of the key outcomes of this research in terms of intellectual merit, significant progress in understanding personalized mobile context was achieved, producing at least one PhD dissertation, one best paper award at a conference, and two journal papers. In terms of the research on mobile misbehavioral context, three journal-equivalent peer-reviewed papers were published at highly selective conferences, as well as software available for commercial companies to license. The research on mobile group context has resulted in one PhD dissertation, and one conference paper, with a submission on the latest mobile group dynamics study results under preparation. Also, a paper outlining the proposed architecture of a context-aware mobile cloud system was published. Parts of this architecture were implemented in the mobile group context prototype, thus making progress towards the goal of implementing an initial prototype of that mobile cloud infrastructure. In terms of broader impact, the research on contextual misbehavior detection has produced licensable software that has proved useful to commercial companies. Further, in terms of diversity, the research was able to support both a female graduate student as well as an undergraduate student in research. The published results of the research are widely available for download on the Web sites of the PIs, co-PIs, and graduate students who participated in the research.
