Research  Projects

Hierarchical Federated Learning Over Wireless Edge Networks: Performance Analysis and Optimization

Federated learning (FL) is revolutionizing machine learning by catalyzing a paradigm shift from cloud-based centralized learning towards distributed, on-device edge learning. FL enables devices to collaboratively train and execute a global learning task by using local processing and simple learning parameters exchange, thus avoiding the communication and privacy concerns associated with sharing large data volumes with a remote cloud. Owing to its attractive privacy, scalability, and communication features, FL will be an integral edge component of Internet of Things (IoT) services such as autonomous systems. However, when deployed over the wireless IoT edge, the performance of FL will be largely constrained by the quality of the wireless links used to exchange the local and global FL model parameters. Since the next-generation IoT will be powered by a wireless cellular system (e.g., 5G), reaping the benefits of FL for the IoT hinges on understanding how wireless factors, such as fading, interference, and delay, impact the convergence and performance of FL (e.g., accuracy, reliability, and convergence time). The goal of this research is to develop a foundational framework that rigorously answers fundamental questions on the achievable FL performance over realistic, large-scale wireless edge networks thus facilitating FL integration unto a real-world IoT. The research is coupled with a well-crafted educational plan that includes a new course at the intersection of communications and machine learning as well as a significant involvement of graduate and undergraduate students at all levels. Broad dissemination and outreach will be ensured via several workshops, tutorials, outreach events, and other tools. This research will develop a novel, holistic framework for performance analysis and optimization of FL over large-scale wireless cellular edge networks. The proposed framework will yield major innovations across both wireless and FL fields: 1) A scalable hierarchical wireless architecture that allows a large-scale implementation of FL over wireless cellular systems, 2) Rigorous performance analysis of hierarchical FL over wireless edge networks that will yield novel FL performance metrics that jointly couple learning performance indicators, such as training accuracy and convergence time, 3) Novel notions of reliability for FL over wireless networks to enable the operation of FL under extreme network conditions and in presence of IoT device mobility and spatio-temporal correlations, and 4) Suitable resource allocation algorithms that can optimize the performance of hierarchical FL over wireless edge networks. The results will be validated using various simulation and experimental means.

Vision-Guided Wireless Communication Systems

Sponsored by: National Science Foundation (NSF) 

Collaborators: Mehdi Bennis (University of Oulu, Finland), Harpreet Dhillon (Virginia Tech)

With the approach of the golden jubilee of the first mobile phone call made in 1973, it is astonishing to see how far wireless technology has come. More recently, the confluence of computing and communications has transformed wireless devices from merely communication devices into powerful computing and sensing platforms. Enabled by these capabilities and equipped with state-of-the-art sensors, cameras, and other non-radio frequency (non-RF) modalities, modern wireless devices are able to simultaneously perform multiple functions including communications, computing, and imaging. In order to exploit synergies across these functions, this project brings together a synergistic US-Finland team with the goal of laying the fundamental science needed to pioneer a novel paradigm of vision-guided wireless system design using which wireless devices can "view" and map their surrounding wireless environment and its features by fusing heterogeneous multimodal information sensed through their RF and non-RF capabilities. Under this new paradigm, wireless network devices can leverage diverse, sensed information about their environment in order to more effectively communicate and compute. This transformative concept contributes towards boosting the performance of future wireless systems (e.g., 6G) thus paving the way for new wireless applications with tangible societal impact, including advanced extended reality, drones, and connected autonomy. The goal of this collaborative US-Finland  project is to develop a novel holistic framework that merges tools from machine learning, distributed optimization, communication theory, and wireless networking to create the foundations of vision-guided wireless networks.

Science of Tracking, Control and Optimization of Information Latency for Dynamic Military IoT Systems

Sponsored by: Office of Naval Research (ONR) - Multidisciplinary University Initiative (MURI) 

Collaborators: MIT, Ohio State University, Virginia Tech (Jeffrey Reed, Tom Hou, and Wenjing Lou)

The goal of this research is to address the military IoT AoI and latency challenges by developing a foundational framework composed of five cogent thrusts that harmonize new notions from information theory, communication theory, signal processing, optimization, control, game theory, learning, and security. This framework enables efficient management of AoI in a dynamic, massive military IoT where the heterogeneous information types and physical world dynamics directly impact how information ages. We start by Thrust I that develops new reinforcement learning, optimization, and information-theoretic approaches to translate the information and physical system dynamics into information whose AoI must be defined using novel multi- mode metrics. This leads to a fundamentally novel latency measure, dubbed multi-mode AoI, that explicitly accounts for the fact that information ages differently in a military IoT. The optimized multi-mode AoI metrics of Thrust I are then refined in Thrust II by quantifying the costs (in terms of overhead latency) for information collection and measurement as well as the value of information. Thrust II will then develop new decentralized optimization and control algorithm to optimize the tradeoff between multi-mode AoI, cost, and value of information. The multi-mode AoI-aware optimization solutions of Thrust II are then extended by Thrust III to massive, self-organizing IoT systems that encompass a very large number of devices, each of which having their own physical dynamics’ constraints, using a fundamentally novel game-theoretic framework. Then, in Thrust IV, the impact of AoI-compromising security threats, such as jamming and replay attacks, is characterized and mitigated. The results of Thrust IV are then used by Thrusts II and III to include security in their objective functions. Finally, Thrust V develops new simulation and hardware testbeds, related to autonomous ground and aerial IoT networks, to validate and demonstrate our framework under real-world IoT constraints.

Towards Highly Reliable Low Latency Broadband (HRLLBB) Communications over Wireless Heterogeneous Networks

Sponsored by: National Science Foundation (NSF) 

Collaborators: Mehdi Bennis (University of Oulu, Finland), Petar Popovski (Aalborg University, Denmark)

Driven by the emergence of the Internet of Things (IoT), next-generation wireless networks will witness a radical departure from the rate-centric designs of yesteryear, toward an ultra-reliable low latency communication (URLLC) paradigm. While current URLLC research has been primarily guided by the need to deliver very short IoT sensor packets, the advent of new IoT applications such as the tactile Internet, is rapidly disrupting this original URLLC premise. Such emerging IoT applications can be classified as highly reliable, low latency broadband (HRLLBB) services as they require joint uplink and downlink transmission of variable-length packets, while guaranteeing high reliability, low latency, and broadband data rates. The goal of this research is, thus, to initiate one of the first concerted US-EU efforts focused on developing the fundamental science needed to seamlessly integrate HRLLBB services into tomorrow's cellular networks. In particular, the proposed research will provide novel analytical tools to facilitate the modeling, design, analysis, and optimization of wireless networks that can cater to HRLLB services. This, in turn, will enable a broad range of novel wireless services with significant societal impacts, ranging from haptics to autonomous systems. 

Towards Surge-Resilient Hybrid RF/VLC Networks

Sponsored by: National Science Foundation (NSF) 

Collaborators: Ismail Guvenc (North Carolina State University)

The successful deployment of smart city-scale wireless services, such as the Internet of Things (IoT), hinges on the presence of a reliable communication infrastructure that can pervasively connect both human and machine-type devices. In order to support the traffic anticipated from an IoT ecosystem, the wireless infrastructure must be resilient to the surge of traffic that can result from a plethora of events that range from hot-spot events and random failures to natural disasters and targeted attacks. The goal of this research is to develop a novel wireless network architecture that leverages both radio frequency (RF) and visible light communication (VLC) to provide resilient and pervasive connectivity. The developed architecture and associated frameworks will bring together novel, cross-disciplinary ideas from communication theory, learning, and game theory to explore the fundamental theoretical and experimental challenges of a hybrid and resilient RF/VLC network. 

Fundamentals of Internet-of-Things with Energy Harvesting and Edge Intelligence

Sponsored by: National Science Foundation (NSF) 

Collaborators: Harpreet Dhillon and Dong Ha (Virginia Tech)

The Internet of Things (IoT) is a massive interconnected ecosystem of wireless connected devices ranging from traditional computing devices to mundane objects. The low-cost nature and limited resources of most IoT devices coupled with the sheer scale of the system introduce fundamentally new design constraints. In particular, a significant fraction of IoT devices will be deployed at hard-to-reach places which makes it challenging to replace their batteries frequently. One appealing solution to this is to allow IoT devices to harvest their own energy from ambient sources, such as the radio frequency (RF) signals of existing wireless networks. To this end, the overarching goal of this research is to develop a holistic framework that will unravel the performance and operational limits of a large-scale energy harvesting IoT system. In particular, the developed framework will provide a unified approach to the modeling, analysis, and optimization of an energy harvesting IoT under realistic operational constraints.

Big Data Analytics for Optimized Planning of Smart, Sustainable, and Connected Communities

Sponsored by: National Science Foundation (NSF)

Collaborators: Virginia Tech (Naren Ramakrishnan, Harpreet Dhillon) and University of Miami (Civil and Architectural Engineering and  School of Architecture)

The goal of this project is to transform villages, towns, and cities into smart, connected, and sustainable communities by developing the first big data-driven holistic approach to joint planning, optimization, and deployment of community infrastructure for systems of critical importance, such as communication, energy, and transportation systems. By bringing together interdisciplinary domain experts from data science, electrical engineering, and civil and architectural engineering, this research will yield several innovations that range from novel big data techniques for faithfully creating spatio-temporal models for smart communities to data-driven performance metrics to explicitly quantify the health of smart communities  and advanced analytical tools  to devise the most effective strategies for deploying, upgrading, and operating various community infrastructure nodes, given the scale, dynamics, and structure of both the data and the community.

Statistical Performance Analysis and Resource Management for Cyber-Physical Internet of Things Systems

Sponsored by: National Science Foundation (NSF)

Collaborators: Harpreet Dhillon (Virginia Tech)

The goal of this project is to develop a foundational framework for the modeling and performance analysis of the Internet of Things (IoT) that will facilitate the management of resources, such as energy and computation, jointly across its cyber and physical realms. By leveraging interdisciplinary tools from stochastic geometry, distributed optimization, and operations research, the proposed framework will yield a number of results that include new statistical models and CPS performance metrics for characterizing the cyber-physical operation of IoT as well as novel distributed optimization algorithms that will adapt the cyber-physical operational state of the IoT devices to the dynamics of the CPS environment, while being cognizant of their stringent resource constraints.

Towards Resilient Smart Cities

Sponsored by: National Science Foundation (NSF)

Collaborators: Virginia Tech (Economics, Computer Science), VTTI, Rutgers University, and FIU

Realizing the vision of truly smart cities is one of the most pressing technical challenges of the coming decade. The success of this vision requires synergistic integration of cyber-physical critical infrastructures such as smart transportation, wireless systems, water networks, and power grids into a unified smart city. Such critical infrastructures have significant resource dependence as they share energy, computation, wireless spectrum, users and personnel, and economic investments, and as such are prone to correlated failures due to day-to-day operations, natural disasters, or malicious attacks. Developing resilient processes that can control and manage these interdependent critical infrastructure resources is therefore a key towards understanding how future smart cities must be operated. The goal of this project is to lay the foundations of resilient smart cities via a coordinated and interdisciplinary approach that relies on machine learning, operations research, behavioral economics, and cognitive psychology and which takes into account both technological and human factors. 

Joint Backhaul and Radio Access Design for Heterogeneous Wireless Networks

Sponsored by: National Science Foundation (NSF)

Collaborators: Harpreet Dhillon (Virgina Tech)

Delivering pervasive access to data-intensive wireless applications is contingent upon enabling wireless cellular systems to sustain the foreseen 1000x increase in the demand for wireless capacity. One promising solution is via wireless network densification in which small base stations are deployed at possible adverse locations, such as lamp posts and the sides of the buildings, to significantly boost the wireless capacity. However, reaping the benefits of such dense cellular networks requires devising novel heterogeneous backhaul solutions that can connect the small base stations to the Internet and core network by smartly and jointly exploiting existing, wired infrastructure, as well as new wireless, possibly in-band, backhaul solutions. The key goal of this project is to introduce a fundamentally new cellular network design framework in which elaborate wireless, wired, heterogeneous, and possibly multi-hop backhaul models are tightly integrated with the access networks to facilitate joint analysis, modeling, and optimization of backhaul and radio wireless access. This proposed framework will marry together notions from stochastic geometry, microeconomics, and wireless communications to enable tomorrow's cellular systems to support bandwidth-intensive wireless applications such as mobile high-definition video streaming, thus expediting their global deployment. 

Energy-Efficient Hyper-Dense Wireless Networks

Sponsored by: National Science Foundation (NSF)

Collaborators: Tohoku University (Japan) and Florida International University

Delivering high-speed, pervasive wireless services to trillions of mobile devices requires a significant transformation to today's wireless cellular networks. Hyper-dense heterogeneous networks (HDHNs), enabled via a viral and large-scale deployment of small cell base stations and mobile devices, are seen as one of the cornerstones of this transformation. However, most existing network design and optimization techniques fall short in handling the scale, density, sustainability, and dynamics of such HDHNs.

This project contributes to the foundations of HDHNs via innovations that include novel mobility state estimation techniques, optimized cell selection and handover methods, and a new optimization framework for self-organizing resource management in energy-efficient HDHNs. The proposed approaches will be evaluated and enhanced via implementation on an experimental testbed using software radio and open source platforms. 

Pervasive Spectrum Sharing for Public Safety Communications

Sponsored by: National Science Foundation (NSF)

Collaborators: University of Nevada - Reno, University of Central Florida, and Florida International University

Next-generation public safety communication (PSC) systems must deliver high-capacity wireless services to public safety personnel and users in disaster-affected areas, with little reliance on infrastructure. Remarkably, modern-day PSC systems have yet to catch up with the past decade's wireless revolution, as they still rely on technologies of yesteryears that fall short on delivering high-speed wireless access. Indeed, coping with the foreseen stringent service requirements in future PSC systems mandates major innovations that can increase spectral efficiency. The overarching goal of this research is to initiate the much-needed leap towards a more open, highly participatory, and pervasive sharing of the wireless spectrum for communication networks, in general, and PSC, in particular. This is done by developing a new framework that tightly integrates the economic and technological challenges of PSC, using both theory and implementation.

A Human-Centered Computational Framework for Urban and Community Design of Resilient Coastal Cities

Sponsored by: National Science Foundation (NSF)

Collaborators: Virginia Tech (Department of Geography) and University of Miami (Civil and Architectural Engineering)

Coastal cities play a critical role in the global economy. However, they are being increasingly exposed to natural hazards and disasters, such as hurricanes, and recurrent flooding due to the rise of sea-levels caused by climate change. The goal of this research is to create new paradigms for the resilient design of urban communities, and uniquely tailored toward the design of coastal cities. Results from this research will help make critical coastal infrastructures more tolerant to damage. The in turn will foster socio-economic resilience by enabling anticipatory interventions. The developed techniques and simulation models will redefine traditional urban design strategies through the integration of architecture, urban design, land-use planning, civil engineering, and advanced computational methods that explicitly consider socio-economic drivers. This project will be conducted in close collaboration with the cities of Miami and Miami Beach. 

Foundations of Social Network-Aware Cellular Communications

Sponsored by: National Science Foundation (NSF)

Collaborators: University of Florida and University of Houston

Providing ubiquitous access to the rapidly expanding suite of mobile social applications has strained modern-day wireless networks, motivating the need for new technologies to optimize the usage of the scarce radio resources. Device-to-device (D2D) communications between mobile devices over the reliable and pervasive cellular network infrastructure constitutes one of the most promising technologies for further accelerating the penetration of mobile social services. As compared to conventional D2D over short-range and limited-capacity technologies such as Bluetooth, D2D over cellular provides longer transmission ranges, improved spectrum sharing, higher capacities, guaranteed quality-of-service, and a broader range of applications and services. Owing to its promising potential, D2D is now viewed by both academia and standardization bodies as a cornerstone technology in emerging 5th generation (5G) wireless systems.

Leveraging cellular D2D for mobile social applications requires addressing a variety of technical challenges that are characterized by a strong interplay between social factors such as content correlation, and wireless features such as network-controlled resource allocation and interference management. The goal of this project is to address these challenges by introducing a novel network optimization framework, cognizant of both social and wireless realms, suitable for ensuring the delivery of high-speed, high-quality social networking services over cellular D2D communications. 

Secure Networked Cyber-Physical Systems

Sponsored by: National Science Foundation (NSF)

Collaborators: Temple University and Florida International University

The goal of this project is to develop novel mathematical principles to design secure and robust networked cyber-physical systems (NCPS) with applications to multiple CPS domains such as power systems and transportation systems. Our goal is to better understand the security of networked cyber-physical systems and more importantly recognize the role of humans, their interaction and decisions, on the security of NCPS. In this regard, a key component of our work is to apply novel stochastic game-theoretic models and control systems theory to better integrate the cyber and physical layers of networked systems while focusing on the role of humans and the effect of their behavior and interaction on the security of the joint CPS. 

Enabling Cellular Networks to Exploit Millimeter-wave Opportunities (NEMOs)

Sponsored by: National Science Foundation (NSF)

Collaborators: Allen MacKenzie (VT), Trinity College Dublin (Ireland), and Queen's University (Belfast, Northern Ireland)

The demand for wireless services continues to increase rapidly, and society has come to rely on the availability of wireless services everywhere and at all times. Conventional wireless services rely on communications at radio frequencies below 5 GHz. These frequency bands are heavily used, and future systems will not be able to meet all demands if operated only with these existing technologies. Recent work has created electronic devices able to generate communications signals at higher frequencies, from 30-300 GHz, which is known as the 'millimeter wave' band. Communications links in these bands have the promise of delivering enormous wireless capacity, but these links behave somewhat more like beams of light than conventional radio communications they can be directional and are easily blocked by objects and by people's bodies. The goal of this project is to provide a holistic analysis of mmWave communications  with a focus on improving services in population-dense environments like stadiums, theaters, and transportation hubs.

Prosumer-Centric Smart Grid Energy Management

Sponsored by: National Science Foundation (NSF)

Collaborators: Rutgers University and Princeton University

The realization of the vision of a smart power grid in which a significant portion of energy stems from renewable sources and other prosumer-owned devices (a “prosumer” is a consumer who can take the dual role of seller and a buyer of electricity) is contingent on large-scale, active prosumer participation in energy management. However, just because such participation can yield significant technological and societal benefits, it cannot be assumed that prosumers will actually become fully involved in the smart grid. Empirical data shows that, despite its exciting prospects, the widespread adoption of the smart grid has been hindered by modest user participation. Motivated by emerging grid scenarios, this project will advance the mathematical framework of prospect theory, a seminal contribution to behavioral economics that won the Nobel Prize, to study grid energy management, as well as to understand and overcome barriers to active user participation in grid energy management.

Machine-to-Machine Communications in Wireless Networks

Sponsored by: Office of Naval Research (ONR) - Young Investigator Program (YIP)

A massive deployment of machine type devices (MTDs) such as wearable sensors, surveillance apparatus, and autonomous vehicles, in wireless networks is foreseen as a major driver of future wireless networking services. Indeed, MTDs will be the cornerstone of many emerging technologies. However, reaping the benefits of MTDs is contingent on the ability of the network to sustain reliable machine-to-machine (M2M) communications. Efficiently integrating M2M communications into wireless networks warrants the introduction of MTD aware resource management mechanisms to optimally allocate scarce network resources (power, time, frequency, space) among MTDs and traditional human-type connections. The goal of this project is to address the M2M challenges in wireless networks by developing a foundational framework for optimized M2M deployment and resource management by weaving together notions from wireless communications, optimization, and network science.

Optimal Placement of Things in an Adversarial Internet of Battlefield Things

Sponsored by: Army Research Laboratory (ARL)

Collaborators: Naren Ramakrishnan (Computer Science, Virginia Tech)

Military battlefields are gradually being transformed into a much anticipated Internet of Battlefield Things (IoBT) ecosystem which will synergistically combine communications, sensing, computing, and autonomy functions. The IoBT will consist of a heterogeneous mix of assets, devices, and systems which must be properly deployed, operated, and managed within a complex environment. One key research challenge facing the deployment of the IoBT is to optimally determine how, when, and where to place (and operate) a heterogeneous number of IoBT “things” that can range from sensors to wearables and autonomous vehicles within the largely adversarial IoBT environment. The goal of this project is to address this challenge by developing the first comprehensive and unified data-driven framework for optimizing the placement of “things” within an adversarial battlefield environment.