COSM-IC International Research Network Interconnect (Seskar, Raychaudhuri)

WINLAB is working with Columbia and CCNY on a large NSF-funded collaborative research project aimed at development of an international networking and wireless testbed by federating US research testbeds including COSMOS, ORBIT, FABRIC and PEERING with experimental facilities in Ireland, Greece, Brazil and Japan.  The federated international testbed will enable experimental research on a wide range of optical, wireless, SDN networking, inter-domain routing and edge computing experiments at a global scale.  This project will further extend international connectivity of network testbeds hosted by WINLAB on the Rutgers campus, and is expected to lead to several research collaborations with groups in Japan, Brazil and Europe on future Internet architecture and protocols.

COSMOS Cloud Enhanced Open Programmable Testbed (Seskar, Raychaudhuri)

The COSMOS project is aimed at design, development and deployment of a city-scale advanced wireless testbed in order to support real-world experimentation on next-generation wireless technologies and applications.  This is a joint project involving Rutgers, Columbia and NYU along with several partner organizations including New York City, CCNY, University of Arizona, Silicon Harlem and IBM.  The COSMOS architecture has a particular focus on ultra-high bandwidth and low latency wireless communication tightly coupled with edge cloud computing.  The COSMOS testbed will be deployed in upper Manhattan and will consist of advanced software-defined radio nodes along with fiber-optic front-haul and back-haul networks and edge and core cloud computing infrastructure.  Researchers will be able to run experiments remotely on the COSMOS testbed by logging into a web-based portal which will provide various facilities for experiment execution, measurements and data collection.

MobilityFirst Future Internet Architecture  (Raychaudhuri, Yates, Seskar)

The aim of this research thrust is to propose and validate a fundamentally new architecture for the Internet as it evolves towards serving an increasing number of wireless/mobile devices. The MobilityFirst project was a collaborative, multi-institutional research initiative under the NSF FIA (Future Internet Architecture) program and its successor program FIA Next-Phase (NP). The goal of the project was to design and validate a clean-slate mobility-centric and trustworthy architecture for the future Internet. The scope of research included specification of the proposed new network architecture, detailed design and verification of key protocol components, analysis of economic and policy aspects, evaluation of network security and privacy, system-level prototyping and validation of the network as a whole, and “real-world” testbed deployments for evaluation by application developers and end-users.  This research has provided considerable visibility for Rutgers in both academic and popular publications, and continues to serve as a foundation for new networking, vehicular, smart city and applications projects.

SA-MCN Standalone Mobile Core Network (Mandayam, Raychaudhuri, Seskar)

This project aims to develop a standalone B5G mobile core network (SA-MCN) integrated with edge computing capabilities.  This project builds on prior NSF-funded basic research on the MobilityFirst future Internet architecture which uses a novel identifier based routing approach to support various mobility requirements including: support for heterogeneous radio access technologies, disruption-tolerant mobility management, multi-homing, multicast, in-network content caching, context-aware message delivery, and service chaining.  The proposed SA-MCN architecture results in a flat mobile core network without the mobility gateways used in the current 4G/5G specification, leading to new mobility service capabilities, significantly lower latency and improved scalability, and improved support for edge cloud computing.

Real-time and Energy-efficient AI (Yuan)

Prof. Bo Yuan’s group is working on the problem of improving AI efficiency in wireless/mobile application scenarios. The state-of-the-art AI technique, especially deep learning, has demonstrated their unprecedented capability and power in various applications. However, deep learning is very computation intensive and storage intensive, thereby posing severe high latency and energy consumption challenge. To democratize AI and enable real-time energy-efficient AI, Yuan’s group has developed a series of high-efficient algorithms and hardware accelerators to improve the speed and energy efficiency of AI.    

Intelligent Physical layer for Wireless Communication (Yuan)

Leveraging the state-of-the-art AI technique, Prof. Yuan and his group is working on next-generation intelligent physical layer technique for 5G and beyond wireless communication. A recent advance is the neural network-enabled channel decoder, which shows significant improvement on both error-correcting performance and decoding speed, and thereby forming a solid foundation for the next-generation high-speed high-reliability wireless network. 

Deep learning for mobile applications (Chen)

The focus is on deep learning methods to run on resource-constrained heterogeneous edge devices (e.g., low-cost sensing platforms, smartphones, and IoT devices) using hardware/software co-design. Along this direction, Dr. Chen has been developing power-centric deep learning accelerators exploiting local possessors of mobile edge devices. Her research includes developing data modeling frameworks for processing extremely large datasets and decomposition algorithms that can effectively decompose deep learning models into blocks according to different processors’ resource characteristics in heterogeneous edge devices, such as GPU, CPU, and low-power processor (LPU). She is also exploring in-memory security mechanisms to secure deep learning models and data for data centers. 

Bridging the gap between model-based and machine learning-based classification in wireless systems (Bajwa, Mandayam)

Hypothesis testing / classification is an important problem in wireless communications, with implications for 5G and beyond wireless systems. In this project, we are devising a classifier that is based on consolidation of data-driven (machine learning) approaches with model-based approaches. These two approaches individually face major challenges towards the fulfillment of the classification task. Specifically, the model-based approach can suffer from our inability to properly estimate the unobservable parameters within a given model. Machine learning approaches, on the other hand, heavily rely on a large number of data with their corresponding labels for the classification task, which might not be feasible. This project is based on the premise that a framework that fuses the model-based approach with the data-driven approach can alleviate the difficulties associated with each stand-alone approach.

Wireless Sensors (Howard)

The focus of this technology area is on improving energy usage and/or exploring novel sensing modalities involving electric, magnetic, RF, or acoustic measurement and control. The cornerstone of the low power program is a line of sensors that are in commercial usage (with a start-up partner) in animal care facilities at Rutgers–reporting environmental conditions for decades using a small coin cell battery. Starting with these sensors, areas we have explored include gas sensors (ammonia and related gases patented), instrumented beds that detect breathing/heartbeat/motion, instrumented pill bottle for compliance, and self-powered foot traffic sensors for occupancy. This group is also working on related projects in sensing that manipulate or detect electric/magnetic/acoustic fields include demonstrating capacitance-based gesture recognition, developing distributed RF phase control for remote powering (patented and license negotiations in progress), and using acoustic phase control to turn ordinary earphones (no modifications) into smart headsets with authentication and a microphone (press release in review).

Agricultural IoT Applications (Howards)

Prof. Howards group recently expanding into the agricultural area by applying sensing technology to apiculture in two areas. First, we have been instrumenting hives with environment sensors and cameras to study Verona mite infestations, a major threat to the commercial pollination industry. We are also in our second season of studies on the bees themselves to get solid information on sensitivity to magnetic and RF fields. They are a novel and easy-to-use model system for these studies and we hope to learn more about the nature of novel biological sensors that have been hinted at in studies of larger and more difficult model systems (e.g. birds).

Federated Learning in IoT Ecosystems and Humans in the Loop (Mandayam) 

Federated learning (FL) is an emerging technique used to collaboratively train a machine-learning model using the data and computation resources of mobile devices without exposing private or sensitive user data. Appropriate incentive mechanisms that motivate the data and mobile-device owner to participate in FL is key to building a sustainable platform. Further, subjective preferences of human decision making is an important consideration that may determine the overall efficacy of such learning. In this project we are studying both incentive mechanisms that promote federation of independent devices and as well as subjective  preferences of humans in the loop of IoT ecosystems. Our work in this project thus far has considered a computation- and communication-efficient method of estimating participants’ contribution levels. The proposed method requires a single FL training process, which significantly reduces overhead. Performance evaluations done using the MNIST dataset, show that the proposed method estimates participant contributions accurately with 46–49% less computation overhead and no communication overhead, as compared to a naive estimation method. Our current work is focused on modeling subjective decision making in FL using prospect theory.

Networked Fusion of Sensor Data for V2X Communication (Gruteser)

  Vehicle-to-vehicle (V2V) communications provides a channel for richer point cloud data sharing, rather than sharing more abstract positions of detected objects. This may improve detection accuracy but the exact benefits remain unquantified. To address this question, this workstream will design and integrate a state-of-the-art point cloud registration method into our sensor fusion simulation pipeline, analyze the benefits for situational awareness, and investigate bandwidth efficient raw data sharing strategies that seek to maximize benefits from raw data (e.g., Lidar) while reducing the amount of information that needs to be exchanged.

Reality-Aware Networks  (Dana, Mandayam)

This NSF project seeks to improve the robustness of wireless sensing and networking technologies through a reality-aware wireless architecture that blends networking and sensing. As wireless sensing and networking technologies make significant strides in today’s world, applications such as automated driving or augmented reality increasingly involve rich sensing of the environment with unprecedented network requirements. To address this, the project develops and studies a reality-aware wireless architecture that blends networking and sensing components, rather than isolating them. Robust perception and high-bandwidth networking benefit innovations across a diverse spectrum of high-impact areas including mixed-reality, robotics, and automated vehicles. 

System Maestro (Ortiz)

An ambient sensing system designed to create Spatio-temporal episodic memory through continuous learning and interaction and to enable a new class of smart sensing applications. The goal of episodic-memory construction is to map signals sensed in the ambient environment (i.e., physical measurements such as a magnetometer, RF, temperature, etc.) to semantically meaningful events over time, thus enabling queries over the activity log of the space. For example, ‘Where did I leave my keys?’ or ‘When was the last time I mopped the kitchen?’. This form of contextual-map building through ambient sensing will enable new sensing applications in smart spaces. We have currently deployed five devices in the WINLAB and 3 in a graduate student’s apartment – with dozens more to be deployed in other campus buildings (currently delayed due to COVID-19). We address the scenario under consideration by designing a fast, accurate, and robust deep-learning-based change-point detection algorithm. The architecture consists of two auto-encoder networks and a joint optimization term based on maximum mean discrepancy (MMD). Our experiments show that the network is sample-efficient, robust to both input distribution and hyper-parameter selection, outperforms the state-of-the-art approaches, and trains 100x–1000x faster, on average.

Project Paz (Ortiz)

Timing interactions are an essential capability of smart environments will be the ability to know when to interact with occupants. We have focused on designing models that can predict when to interrupt the driver of an active vehicle in Project Paz. While our experiments do not occur in an office or home setting, we conjecture that the lessons learned in this environment – with dynamically changing context, emotion, and physiological state – will inform active agents’ design in less dynamic contexts than those in a home or office setting. Our model achieves an improvement of 24% over the state-of-the-art. Also, we train the model to avoid false positives – determining that it is a good time to interrupt when it is not – since these may result in deactivation by the driver. The data also show that most uninterruptible times are short and that looking ahead can improve model accuracy, currently being explored in ongoing work.

Mobile Healthcare (Chen)

Many of her research projects in mobile healthcare utilize WiFi signals to provide non-intrusive and passive wellbeing monitoring without requiring the user to wear any devices or participate actively, which are especially useful in elder care communities. She is also exploring practical and effective solutions equipping the eldercare communities for greater preparedness and resilience to pandemic situations such as COVID-19 leveraging wireless signals and multi-modal sensing technologies in Internet of Things (IoT). 

Engineering and Controlling the Immune System (Trappe)

The immune system is a complex system of many agents interacting and communicating with each other through chemical signaling processes. By modeling the complex dynamics underlying the immune system and pathologies such as viral infections and cancer, it is possible to formulate immunological control problems using the connections between biology and DSP, machine learning, communications, networking and graphs, and control theory. The goal of this project is to arrive at mathematical and analytical tools that will support the next generation of applied pharmaceutics, with the aim of supporting personalized medicine and immunotherapies capable of adapting to the specifics of individual patients more rapidly than is currently possible. 

Cybersecurity (Chen)

One active research project is using the cross-domain speech similarity between the audio and vibration domains to provide enhanced security to the voice assistant systems (e.g., Google Home and Amazon Alexa). She has developed a training-free user authentication system that can verify users’ voice commands via the users’ unique vocal traits across the audio and vibration domains by exploiting low-cost sensors on wearable devices.

Foundations of Privacy-Preserving Distributed Learning (Sarwate)

This project seeks to develop principles and algorithms for differentially private learning and inference algorithms that operate on distributed or decentralized data. A prime motivating example are collaborative research systems in which several parties (e.g. medical research groups) wish to collaborate to learn jointly from data held by each party.

Privacy-Enabled Resource Management for IoT Networks (PERMIT) (Sarwate, Mandayam)

Large-scale sensing systems raise important privacy challenges in data collection and communication. This project uses privacy metrics originally developed in the context of databases to analyze communication protocol. The ultimate goal is to understand how privacy resources should be managed in these emerging networked systems.

Secure IoT (Wu, Bajwa, Mandayam)

The NSF SARE (Secure Analog-RF Electronics and Electromagnetics) EAGER project aims to address a critical security issue in IoT applications that are susceptible to malicious spoofing attacks via an innovative physical-layer (PHY) solution combining non-contiguous orthogonal frequency division multiplexing (NC-OFDM) transmission and a directional modulation retrodirective array. As compared with traditional OFDM transmissions, NC-OFDM transmissions take place over a subset of active subcarriers to either avoid incumbent transmissions or for strategic considerations. As such, NC-OFDM transmissions have low probability of exploitation characteristics against classic attacks based on cyclostationary analysis. On the other hand, retrodirective antenna arrays are well known to be able to respond to an interrogator by sending signals back to the interrogator location without a priori knowledge, which is particularly useful in a multipath-rich environment. By incorporating the directional modulation technique, the antenna array will corrupt the information by distorting the digital modulation’s constellation diagrams in all unwanted transmitting directions. One way to realize the directional modulation functionality is to use time-modulated antenna arrays, in which the aliasing effects resulting from the time-modulation frequency are used to distort the signals in the undesired directions. Furthermore, the unique integration of NC-OFDM and directional modulation enabled by a time-modulated retro-directive antenna array whose modulation frequency is the NC-OFDM subcarrier can potentially lead to an unprecedented level of PHY hardware security against spoofing attacks by an adversary, even when the adversary is equipped with sophisticated machine learning-based attack techniques.

Ultra-Wide Band based positioning and ranging for railways (Spasojevic)

We are conducting a study (jointly with CAIT) to understand the promising use cases of the UWB technology in the rail industry and develop an innovative solution to leverage the positioning capability for UWB for locating and tracking rail-related critical objects (e.g., trains, maintenance of way vehicles). Specifically, the research is focused on optimizing the UWB sensor in-field deployment, improving the location accuracy and system robustness.

Post-Correction Techniques for Linearization of Wideband Analog-to-Digital Converters (Spasojevic)

Inherent non-linearities in ADC design and manufacturing create undesirable artifacts which reduce Signal-to-Noise Ratio (SNR) and worsen the Spurious-Free Dynamic Range (SFDR) of the ADC. Wideband ADCs are particularly susceptible to this, as they often have lower quantization resolution which makes these artifacts more prominent. Our research seeks to develop post-correction mechanisms for the non-linearities in such wideband, low-resolution ADCs to improve their SNR and SFDR.

Improving LPD of QS-DS-CDMA Via Cyclostationary Feature Reduction (Spasojevic)

To ensure easily and flexibly deployable ad-hoc communication networks via a reduced coordinating infrastructure, we aim at reduced synchronization requirements.  In military and enhanced privacy settings, reduced communication detectability is also of crucial importance. Imperfect synchronization can be dealt with by making use of Quasi-Synchronous Code Division Multiple Access (QS-CDMA) signaling, while low probability of detection (LPD)  can be achieved by employing randomization techniques that disrupt the inherent signal structure. 

Enabling Spectrum Coexistence between 5G and Passive Weather Sensing (Mandayam, Wu)

The 5G band allocated in the 26 GHz spectrum referred to as 3GPP band n258, has generated a lot of anxiety and concern in the meteorological data forecasting community including the National Oceanic and Atmospheric Administration (NOAA). Unlike traditional spectrum coexistence problems, the issue here stems from the leakage of n258 band transmissions impacting the observations of passive sensors (e.g. AMSU-A) operating at 23.8 GHz on weather satellites used to detect the amount of water vapor in the atmosphere, which in turn affects weather forecasting and predictions. In this project, we study the impact of 5G leakage on the accuracy of data assimilation based weather prediction algorithms by using a first order propagation model to characterize the effect of the leakage signal on the brightness temperature (atmospheric radiance) and the induced noise temperature at the receiving antenna of the passive sensor (radiometer) on the weather observation satellite. We are currently working on future directions for both improved modeling of 5G leakage effects as well as mitigation using cross-layer antenna techniques coupled with resource allocation. Our results to date have been presented at the IEEE 5G World Forum in 2020.Our work has also received extensive coverage in both national and international media.

SII Center: IRIS: Interdisciplinary National Research Center for Innovations in Spectrum

This project (which is a joint effort by Rutgers, Columbia, NYU, Princeton, Arizona, UT Austin, U Syracuse and Oregon State) is aimed at the creation of IRIS: Interdisciplinary National Research Center for Innovations in Spectrum. The IRIS center will conduct both fundamental and applied research on a comprehensive set of technical, economic and policy challenges associated with efficient, abundant and secure use of wireless spectrum, and will also serve as a national resource for spectrum related open source technologies, experimental infrastructure/methodologies and best practices across industry, government and academia. The IRIS center will also incorporate frameworks for extensive research collaboration with industry and government, international partnerships, and comprehensive educational outreach, workforce development and broadening participation programs.

The IRIS center is inspired by the proposing team’s shared vision of a holistic new approach to spectrum based on the four foundational principles of explore, conserve, quantify and share, which when implemented together will lead to an abundance of efficiently used spectrum sufficient to meet the demands of emerging wireless applications for decades to come. Specifically, the solution to spectrum scarcity that we are advocating is based on: (1) significantly adding to spectrum supply through discovery of new frequency bands currently not in use (“exploring spectrum frontiers”); (2) increasing spectrum efficiency by an order of magnitude via fundamental advances in radio and interference mitigation technologies (promoting “spectrum conservation”); (3) declaring, observing and measuring spectrum use to enable dynamic resource allocation, promote accountability and protect authorized users (“quantifying spectrum use”); and (4) introducing a new information architecture for spectrum that supports a range of cooperative sharing algorithms and market transactions (“data-driven spectrum sharing”). This cross-layer data-centric approach to spectrum management will be realized in a focused multidisciplinary center that promises important benefits to both commercial communications and passive scientific use cases such as radio astronomy and satellite sensing.