End-to-end delay is an inherent important metric of all modern communication systems. As the end-to-end delay requirement is becoming more stringent (5G wireless technology is aiming for a latency of less than one millisecond) and as the improvement of local delay is quickly approaching the physical limits of the individual components, it is of paramount interest to study the fundamental limits of end-to-end delay from a network's perspective. Specifically, what is the largest throughput one can transmit over a multihop network while respecting a small end-to-end delay requirement? The research in this project will uncover new delay-aware fundamental limits of network communication and computation. Such results will have a basic impact on information theory, algorithmic graph theory, and communication complexity theory and promote bridging and sharing across technical disciplines that are normally studied by different scientific communities. The design of low-delay schemes will also have immediate and long-standing impact on many practical applications, including Internet of Things, cyber-physical systems, video conferencing, and mobile gaming. The project also includes plans for integrating the research into graduate-level lecture materials.
The research in this project is centered around answering the question ?what is the information-theoretic capacity of a networked system under a stringent (small) finite end-to-end delay requirement??. The distinct nature of this problem requires a cross-disciplinary approach spanning graph-theory, computer science, and information theory. This project develops various new tools including new traffic models, new probabilistic frameworks, and new distributive functional computation paradigms and provides a long overdue theoretic study that explores both the graph-theoretic and information-theoretic aspects of perishable network information flow. The research is divided into two thrusts: one focused on end-to-end delay-constrained multihop network capacity and the other focused on delay-constrained distributed function computation.
