This is a joint project between UCLA, Yale University and University of Minnesota funded by the US National Science Foundation (NSF) under the CNS award 2146838.
Synopsis
This project aims at developing a framework that enables the characterization of fundamental resilience-capacity trade-offs over multi-hop mmWave networks, by leveraging novel opportunities such as beam scheduling, novel coding techniques as well as learning-based methods. Developing fundamental resilience-capacity performance trade-offs for mmWave networks, and offering guidelines on how to achieve these at scale with low-complexity techniques, are not well explored areas; this project aims to address this gap. This project brings together tools from several disciplines; in particular, we combine tools from information and coding theory, machine learning, algorithms, and networking to address the proposal goals.
Goals
Overall, the major goals in this project can be summarized as follows:
Design mechanisms that suitably distribute the traffic across multiple paths in the mmWave network, with the two-fold objective of ensuring resilience against link blockages and/or node failures and achieve high end-to-end packet rate delivery.
Create (proactive) resilient information coding schemes that exhibit a graceful degradation in performance despite (a priori unknown) link blockages and node failures.
Develop low-complexity communication and scheduling mechanisms with high-degrees of resiliency at-scale that can be deployed over dense multi-hop mmWave networks where network resources might need to be shared to accommodate several users.
Develop intelligent mechanisms for multi-hop mmWave networks that leverage feedback to rapidly identify link blockages and node failures, and gracefully adapt to these.
Design decentralized backpressure algorithms that adapt to network failures.
-
Leverage Reinforcement Learning (RL) techniques to autonomously adapt to failures.
-
Analytically compare the proactive and reactive strategies across a range of performance metrics. Develop hybrid strategies that combine the two approaches, and provide network synthesis guidelines.
-
Develop a proof-of-concept and demonstrate the performance of proactive, reactive and hybrid approaches of the previous tasks by means of testbed experimentation.
-
Evaluate the performance of our mechanisms over more complex networks and traffic configurations than those that we will experimentally deploy.
Contact
Prof. Christina Fragouli UCLA
Prof. Leandros Tassiulas Yale University
Prof. Suhas Diggavi UCLA
Assistant Prof. Martina Cardone University of Minessota
|