CAREER: Vision and Learning Augmented D-Band Networking and Imaging

Project Overview

Synopsis: Millimeter-wave (mmWave) is the core wireless technology to enable new applications in transportation, entertainment, education, and telemedicine. Specifically, the recent availability of inexpensive hardware above 100 GHz makes the time ripe for bringing D-band (110-170 GHz) mmWave networks to the masses. However, D-band mmWave networks bring new challenges in optimizing the deployment of picocells, coordination and adaptation of mobile links with unprecedentedly wide frequency options, and a disruption-free confluence of networking-imaging. This research project addresses these key challenges and improves the performance, reliability, and usability of mobile D-band networks. The project will design machine learning augmented scalable D-band systems and networks, and integrate them into applications, such as Augmented Reality (AR), drone delivery, and autonomous cars. The research outcomes will impact the broader population by: (1) bringing ubiquitous and high-quality bandwidth to underserved users; (2) enabling efficient use of spectrum to better utilize this nationally important resource; and (3) elevating the utility of networking devices by enabling several critical applications on them. The proposed research will be disseminated through publications, open-source software and datasets, and close collaboration with industry partners. It will be integrated into education by designing new undergraduate and graduate cross-disciplinary wireless curricula and involvement in broader community outreach activities.

This project aims to enable the practical adoption of D-band mmWave networks and applications by solving the fundamental challenges in deployment, link adaptation, coordination, and unified networking-imaging. Specifically, the project explores an optical vision and deep learning augmented paradigm by thoroughly understanding the physical properties of the D-band channel, building measurement-driven empirical and learning models, and designing practical, real-time systems. Successful execution of this project would enable the following. (1) A framework for optimal deployment and a ?what-if? analysis tool to help optimize the cost and benefits of D-band deployment in both indoor and outdoor environments. (2) Link adaptation and coordination protocols that significantly minimize latency and maximize throughput and efficiency for scalable D-band networking. (3) A unified networking-imaging protocol that reduces disruptions to the throughput and latency and overcomes challenges with the channel specularity to enable high-resolution D-band images. The project will design, build, and empirically validate the proposed systems in a D-band testbed, and the testbed will be extended into an educational platform that enhances the knowledge of wireless networking and sensing for students at different levels.

Publications and Other Products

  • MiShape: Accurate Human Silhouettes and Body Joints from Commodity Millimeter-Wave Devices
    Aakriti Adhikari, Hem Regmi, Sanjib Sur, Srihari Nelakuditi
    IMWUT’22 Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, Atlanta, USA, September 2022.  [Paper] [Slides] [Talk]
  • Towards Deep Learning Augmented Robust D-Band Millimeter-Wave Picocell Deployment
    Hem Regmi, Sanjib Sur
    SIGMETRICS/PERFORMANCE’22 Proceedings of the ACM Joint International Conference on Measurement and Modeling of Computer Systems, Mumbai, India, June 2022. [Paper] [Poster]
  • D3PicoNet: Deep Learning Networks for Robust Deployment of D-Band Millimeter-Wave Picocells
    Hem Regmi, Sanjib Sur
    WoWMoM’23 IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks, Northeastern University, Boston, Massachusetts, June 2023. [Paper] [Slide] [Talk]
  • Argosleep: Monitoring Sleep Posture from Commodity Millimeter-Wave Devices
    Aakriti Adhikari, Sanjib Sur
    INFOCOM’23 IEEE International Conference on Computer Communications, Stevens Institute of Technology, New York Area, May 2023. [Paper] [Slide] [Talk]
  • Towards Robust Pedestrian Detection with Roadside Millimeter-Wave Infrastructure
    Hem Regmi, Vansh Nagpal, Sanjib Sur
    INFOCOM’23 IEEE International Conference on Computer Communications, Stevens Institute of Technology, New York Area, May 2023. [Paper] [Poster]
  • MatGAN: Sleep Posture Imaging using Millimeter-Wave Devices
    Aakriti Adhikari, Siri Avula, Sanjib Sur
    INFOCOM’23 IEEE International Conference on Computer Communications, Stevens Institute of Technology, New York Area, May 2023. [Paper] [Poster]

Education, Outreach, and Other Broader Impacts

  • Aakriti presented the paper “MiShape: Accurate Human Silhouettes and Body Joints from Commodity Millimeter-Wave Devices” at the ACM Ubicomp 2022
  • Aakriti received the travel grant award to attend ACM HotMobile 2023 and IEEE INFOCOM 2023 conference
  • Aakriti is invited to participate in the CRA-WP Grad Cohort for Women, 2023
  • Zhuangzhuang received the travel grant award to attend ACM/IEEE IPSN 2023 conference
  • Hem received the travel to attend the ACM SIGMETRICS 2022 conference
  • Initial research training for Zhuanzhuang Gu, a Ph.D. student at USC
  • Topics integrated as a part of an advanced, graduate-level IoT course. High excitement among students, with several intending to take this up as a longer-term research project
  • Topics are integrated into a summer EE high school camp focused on wireless applications
  • PI Sur presented at Samsung Research, 5G/6G Mobility Innovation Lab, March 2023
  • PI Sur is organizing the third version of the IEEE STEERS workshop, in conjunction with IEEE/ACM CCGRID 2023.
  • PI Sur served as the poster and demo co-chair for ACM HotMobile 2023.
  • PI Sur served as the publicity co-chair for IEEE COMSNETS 2023
  • PI Sur served as the judge for Senior division projects competition in USC Science and Engineering Fair in March 2023
  • PI Sur developed and taught a Graduate IoT class with 5G Millimeter Wave: Wireless and Mobile Systems for the IoT in Spring’23

Open-Source Software and Data Release

  • 122 GHz measurement datasets for indoor environment: Coming soon.
  • Millimeter-wave reflections for different sleep postures: Coming soon.