AI for Next-generation Self-driving networking

    Software defined networking (SDN) has been widely envisioned to be the next-generation networking paradigm for both wired and wireless networks (e.g., Google B4 SDN data center networks and AT&T SDN cellular systems). The main ideas of SDN are (i) to separate the data plane from the control plane; and (ii) to introduce novel network control functionalities based on an abstract representation of the network. The SDN technologies can lead to extremely flexible computing, networking, and storage infrastructure, which dramatically improve network resource utilization, simplify network management, reduce operating cost, and promote innovation and evolution. In the past, we are (1) developing scalable and efficient traffic engineering solutions (including control traffic balancing, optimal network planning, traffic QoS classification, and SDN routing), (2) designing novel SDN architectures for wireless systems such as 5G cellular systems and next-generation underwater networks, and (3) implementing and prototyping custom-designed software-defined wireless mesh network (http://www.softmeshnet.com/). Those solutions aim to counter the fundamental scalability limitation of current SDN technologies, while keeping the benefit of network programmability and virtualizability for globally optimal computing and networking performance. Recently, my research focuses on developing AI-oriented SDN architecture for next-generation self-driving networks. Our ultimate goal is to show that the coupling of the programmable control and programmable measurement of SDN with the inference capabilities of reinforcement learning and deep learning promises unprecedented opportunities to realize self-driving self-managing computer networks, which make network management and control decisions in a real-time and automated fashion. [See More Project Details Here]

    With the Support of National Science Foundation.

Distributed and Federated Deep Learning

    Federated learning (FL) has emerged as a key technology for enabling next-generation privacy-preserving AI at-scale, where a large number of edge devices, e.g., mobile phones, collaboratively learn a shared global model while keeping their data locally to prevent privacy leakage. Enabling FL over wireless multi-hop networks, such as wireless community mesh networks and wireless Internet over satellite constellations, not only can augment AI experiences for urban mobile users, but also can democratize AI and make it accessible in a low-cost manner to everyone, including people in low-income communities, rural areas, under-developed regions, and disaster areas. The overall objective of this project is to develop a novel wireless multi-hop FL system with guaranteed stability, high accuracy and fast convergence speed. This project is expected to advance the design of distributed deep learning (DL) systems, to promote the understanding of the strong synergy between distributed computing and distributed networking, and to bridge the gap between the theoretical foundations of distributed DL and its real-life applications.

    With the Support of National Science Foundation.

AI-enabled Human Digitalization and Applications

    Recovering 3D human models from monocular images/vidoes has drawn more attention in recent years due to vast practical applications that require 3D human modelings, such as gaming, human-computer interaction, and virtual systems. We aim to develop a set of deep learning-based models to enable fast and high-quality human digitalization and enable a set of downstream applications that exploit digital 3D human models, such as digital arts/filming, gaming, sports analysis, health care, and surveillance

    With the Support of National Science Foundation.

Deep Learning for Smart and Secure Wireless Sensing

    Millimeter wave (mmWave) communication is considered to be one of the key enabling technologies of next generation very high throughput wireless networks. Currently, we are developing mmWave device forensics technologies, which aim to identify unique mmWave fingerprints of unauthorized mmWave devices and the security attacks launched by these devices [J1, J4, C8]. Moreover, mmWave sensing is an emerging non-intrusive sensing technology with the advantages of high-resolution motion capture. long-range detection, independence of lighting conditions, ability to penetrate materials (e.g., clothing and wall), excellent performance under adverse environmental conditions (e.g., smoke, fire, rain, fog, dust and snow), and inherent privacy-preserving capability. Currently, we are developing deep mmWave sensing technologies, which exploit deep neural networks to automatically extract the salient and representative features from rich and high-dimension mmWave signatures, such as micro-Doppler signatures and 3D point clouds. Our objective is to achieve simultaneous tracking, action recognition, pose estimation and user identification at-scale. Our deep mmWave sensing technologies will promise many applications in smart surveillance, smart manufacturing, health care, human-computer interfacing, and VR/AR/MR..

    With the Support of Departemnt of Energy and National Science Foundation.

Swarming Cyber-Physical Systems

    Swarming cyber-physical systems consists of a collection of mobile networked agents, e.g., underwater autonomous vehicles and drones, which collaboratively accomplish a common mission by exploiting their on-board sensing, computing, communication, and locomotion capabilities. The inherent mobility, adaptability, cooperativeness, and robustness of swarming cyber-physical systems have promised a wide range of civilian and military applications. Despite its promising applications, the realization of swarming CPS is facing fundamental challenges because mobile computing entities, e.g., robots, need to collaboratively interact with phenomena of interest at different physical locations, where the contextual surroundings of those locations exhibit inherent uncertainties, e.g., unreliable communication channels, unpredictable mobile obstacles, and prevailing dynamics of the phenomena of interest. Such unique challenge motivates our current research on the development of novel sensing-communication-motion co-design solutions to seamlessly integrate collaborative sensing, distributed computing and wireless communication with mobile observations about physical world. Currently, we are developing the first fully-autonomous swarming cyber-aquatic system in the literature, which is realized through a network of intellgient underwater robots that are constituted by AI-powered autopilot system, bio-inspired swarming intelligence, Magnetic-Induction (MI) based underwater transceivers, innovative 3D localization module, underwater wireless recharging technologies and high-performance underwater networking protocols.

    With the Support of National Science Foundation.

UAV Networks

    A critical lesson learned from every disaster, regardless of its size or the level of devastation, is the importance of communications for managing and coordinating the response, maintaining the rule of law, and keeping the public safe and fully informed. Unfortunately, during the first 72 hours of a response, communications may be partially or completely disrupted due to damaged facilities, widespread power outages, and lack of access by restoration crews and equipment to the impacted area. To address such challenge, I am developing a fast deployable and high-speed communication system, namely SKYNET, which uses a swarm of small unmanned aerial vehicles (UAVs), e.g., quadcopters, to collaboratively and rapidly establish a reliable, resilient, cost-effective, and high-speed communication backbone. Our current research efforts focus on prototyping the aerial small cell communication systems, developing the effective air-to-ground channel models, proposing the joint motion and communication optimization solutions, and investigating the effective energy management solutions.

    With the Support of John A. See Innovation Award.

Internet of Nano-Things

    Nanotechnology is enabling the development of miniature devices able to perform simple tasks at the nanoscale. The interconnection of such nano-devices with traditional wireless networks and ultimately the Internet enables a new networking paradigm known as the Internet of Nano-Things (IoNT). Despite their promising applications, the peculiarities of nano-things introduce many challenges in the realization of the IoNT. First, even with the advanced grapheme-based nano-transceivers and nano-antennas, nano-devices have to operate in the Terahertz (THz) bands, which suffer from a very high propagation loss, while providing a very large bandwidth. Second, the very limited amount of memory equipped on the nano-devices may only allow one packet to be temporally queued before being transmitted. Third, the nano batteries can only hold very limited amount of energy and it is infeasible to manually replace them and recharge them. My research aims to address the above mentioned challenges by developing radically new and computation-light PHY/MAC/NET layer solutions for bufferless nano-devices, which can maximize network throughput, while achieving perpetual operations by jointly considering the energy consumption of THz communications and energy charging with piezoelectric nano-generators.