Communication and storage networks play an integral role in various aspects of modern life. The design of an efficient infrastructure of communication and storage relies significantly on algorithms for efficient information flow among its various components. The multiple unicast network capacity characterization problem is an information theoretic abstraction that captures the fundamental challenges in information flow in several communication and storage systems. Despite its importance, development of optimal codes in multiple unicast networks is a challenging unsolved problem in information theory.
Recently the idea of interference alignment, which was originally discovered in the context of wireless systems, has been shown to be an important tool in multiple unicast network coding. However, uncovering interference alignment based achievable coding schemes remains an art; alignment-based solutions typically must be carefully handcrafted based on the specific network scenario under consideration. A dearth of systematic alignment based approaches has restricted the impact of this idea to a narrow class of network topologies and configurations. The objective of this project is to fill this gap by developing systematic interference alignment based algorithms for multiple unicast network coding. The project will include the study of several toy models of networks, which are tailor-made to expose the role of alignment. Insights obtained from the toy models will be incorporated into network coding algorithms for general networks over directed acyclic graphs. The performance of the algorithms developed will be studied through analytical and numerical methods.
The proposed work involves fundamental contributions to the field of network information theory, specifically to the sub-topic of network coding, through the study of canonical problems in the field. The developments of the project will impact the design of data centers, content distribution networks and wireless communication networks. The research developed will be disseminated through peer-reviewed conferences and journals. The source code for algorithms developed will be made available to the public.
|Effective start/end date||8/1/15 → 7/31/18|
- National Science Foundation: $174,371.00