CIF: Small: Timing Optimization Over Random Network Asynchrony - Theory And Distributed Algorithms

点击翻译按钮获取中文摘要

A key force driving technology revolutions in the 21st century is the ever accelerating deployment of more powerful but physically smaller computing/sensing devices that are constantly and interactively connected through newer-generation wireless networks. There is interest in supporting device densities of up to one million devices per square kilometer, which, once achieved, would enable exponentially many innovations that further transform the landscape of modern society. A grand challenge of supporting this massive scale of machine-type communication is that modern wireless networks, while offering high throughput and broad coverage, are inevitably susceptible to random delay. This delay results in a phenomenon called "network asynchrony," for which any message sent by a device is always outdated to some random degree when it actually arrives at its intended destination. As such, the local views of any two nodes regarding network information are always "out-of-sync," and the key challenge is how each individual node can optimally communicate and collaborate with others despite their asynchronous local views. This project investigates the optimal network collaboration policies under network asynchrony, with results potentially leading to a high-performance design paradigm for the much-needed next-generation machine-type communication protocols, substantially improving the communication efficiency of autonomous vehicles, sensor networks, and many other Internet-of-Things devices.Motivated by the recent discovery of data freshness control, this projects studies how sensors and controllers, and the entire network in general, can optimally collaborate over a temporally noisy information loop, and particularly how to perform (transmission) timing optimization over random network asynchrony. Three major thrusts will be investigated. In thrust 1, a new fixed-point equation framework crystallizing the existing theoretic developments will be explored, opening up new venues for optimal scheduling characterization and numerical evaluation. In thrust 2, new distributed algorithms and data-freshness-control schedulers that optimally adapt to any unknown delay distribution will be developed. In Thrust 3, joint consideration of acknowledgement-centric, acknowledgement-free, and hybrid designs. and quantifying the impact of transport-layer design choices on data freshness control will be investigated. The results of these thrusts will lead to new theoretical characterizations, distributed algorithms, and protocol designs for autonomous machine-type communications in a way similar to the development of TCP-based flow-control algorithms in the early days of Internet protocols.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.