Simplifying Large-Scale Complex Networks
ECE Assistant Professor Milad Siami was awarded a $300K NSF grant for “Sparse Sensing, Actuation, and Communication in Complex Networks.” The project aims to make large-scale complex networks simpler by sparse interactions in the right place at the right time.
Abstract Source: NSF
Social networks that connect people, smart power grids, pandemic influenza mitigation, and flying drone display are all examples of large-scale complex networks. Due to the various worldwide applications of these networks, achieving efficient and robust control and monitoring of such networks is crucial, especially using limited resources. We seek to accelerate the development of novel technologies and advanced algorithms for complex dynamical networks using recent advances in theoretical computer science, graph theory, and machine learning. The results of this project will provide a rich set of building blocks to find a sparse yet important subset of available sensors and actuators in the complex network to enable precise control and high-resolution monitoring of the entire network. As an integral part of this research program, we also propose an educational plan involving K-12, undergraduate, and graduate-level education. The outreach element will improve the pre-college students’ awareness of the impact and attractiveness of both research and engineering careers. Moreover, we aim to leverage the resources at Northeastern’s University Program in Multicultural Engineering to broaden students’ professional skills, find valuable contacts, and explore one or more career paths. We also expect to collect the results of this research in a special topic graduate-level course documenting the theoretical underpinnings and algorithmic implementations of sparse interaction in complex networks.
Given the increasingly large-scale nature of complex networks, it is crucial to rapidly estimate and control the state of the overall network in a distributed fashion with provable performance guarantees, while using a minimum amount of actuations, observations, and communications at each node and link over time. The main goal of this proposal is to generalize collaborative subset selection methods, and adaptive samplings in machine learning to (i) reduce network complexity and data overload by sparse scheduling of available Sensor/Actuator/Coupling measurements in complex networks, (ii) deal with structured uncertainties, missing information, and corruptions on the update dynamics of each agent or communication link, and (iii) handle a large class of controllability and observability performance measures that are not supermodular or submodular.
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.