Project Details
Description
An increasing number of applications, including surveillance, medical monitoring, automatic translations and gaming, rely on the capability of mobile wireless devices to carry out computation-intensive tasks in a timely manner. This requirement conflicts with the expectation that mobile devices should run on a battery without needing frequent recharging. A promising solution to this challenge is mobile cloud computing, that is, the offloading of computation-intensive tasks to a cloud service provider by means of wireless transmission. However, the energy and latency entailed by wireless transmission may offset the potential gains of mobile cloud computing. This project proposes to tackle the outlined problem via the development of effective, low-complexity, scalable and flexible offloading strategies that operate over a mobile fog computing architecture, in which small-cell base stations are endowed with computing capabilities to offer proximate wireless access and computing. The insights gained from the successful completion of this project will be beneficial for a gamut of other exciting problem domains that require large-scale optimization, including big data mining, signal processing, machine learning, and smart grid. The research agenda is complemented by a multidisciplinary educational plan that targets both undergraduate and graduate students via hands-on learning and experimentation activities. Industrial collaboration is also envisaged through internship and co-op opportunities.
The inter-layer optimization of the computation and communication resources in a mobile fog computing network yields unstructured nonconvex mixed-integer problems, which are unexplored and challenging, and whose formulation depends on whether the mobile applications are splittable, i.e., divisible into subtasks that can be individually offloaded, or not. Since the problems at hand do not lend themselves to the application of existing iterative optimization techniques, such as Difference-of-Convex programming, a class of scalable and flexible solution methods with controllable convergence, complexity and overhead is introduced based on a novel successive convex approximation framework. In the case of splittable applications, the analytical and algorithmic framework is augmented by the application of message passing strategies that leverage the call graph representation of the mobile applications.
The inter-layer optimization of the computation and communication resources in a mobile fog computing network yields unstructured nonconvex mixed-integer problems, which are unexplored and challenging, and whose formulation depends on whether the mobile applications are splittable, i.e., divisible into subtasks that can be individually offloaded, or not. Since the problems at hand do not lend themselves to the application of existing iterative optimization techniques, such as Difference-of-Convex programming, a class of scalable and flexible solution methods with controllable convergence, complexity and overhead is introduced based on a novel successive convex approximation framework. In the case of splittable applications, the analytical and algorithmic framework is augmented by the application of message passing strategies that leverage the call graph representation of the mobile applications.
Status | Finished |
---|---|
Effective start/end date | 9/1/15 → 8/31/18 |
Funding
- National Science Foundation
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.