Limits of Information Flow in Networks of Two-Way Channels

Texas A&M Engineering Experiment Station, College Station TX

Investigators

Abstract

Communication networks are critically important in the dissemination of information in today's technological society. Several academic disciplines have undertaken the study of various aspects of their operation, but an integrated theory of networking has been elusive. The objective of network optimization theory is to most effectively use network resources, but recent work in network coding has demonstrated that a standard assumption of flow conservation from this theory has imposed artificial restrictions on the workings of communication networks. Network information theory considers more general models of communication and offers some bound on the limits of information transfer in a communication system involving multiple senders and receivers. This research fuses aspects of network optimization theory and network information theory in order to find and specify new limits of information transfer across a communication network. The investigators use the two-way channel from information theory as a general model of the transmission medium between neighboring processors of a communication network and address fundamental optimization problems pertaining to the analysis and design of networks of two-way channels. The investigators seek optimal routing algorithms in multicast settings and a broad classification of situations where network coding can or cannot improve upon routing. The investigators further aim to categorize network scenarios in which a bidirected cut set bound derived using information theoretic arguments is tight and to explore a network information counterpart to Kirchoff's law when the bidirected cut set bound is not tight. The research also involves extending the notion of minimal cost flow as an initial approach to treating quality-of-service issues and devising networks with least cost or capacity that meet requirements on information transfer.