The research training in communications, networks, and security areas focues on creating NextG wireless systems with enhanced communication performance, novel sensing functionalities, strong attack resiliency, and high energy efficiency, for uses in Augmented, Virtual or eXtended Reality (AR/VR/XR), autonomous driving, massive interactive real-time applications, advanced industrial, manufacturing, and telehealth. Specifically, we design wireless systems and algorithms to control or explore the wireless propagation characters, thus improving signal quality, achieving ubiquitous, low-power sensing, and preventing or reducing attack vectors, such as jamming and eavesdropping.

Wireless Sensing for Physiological Monitoring. The NextG wireless vision create precise representations of the physical world using wireless signal calls for new sensing solutions that extend beyond traditional radar functions of localization, tracking, and object recognition. Wireless sensing using the mobile communication network as a sensor (with the evolution of cellular systems to mmWave bands and sub-THz bands, deployed in small cell configuration) has the potential to become a key component of NextG wireless systems. However, the challenge is to achieve sub-wavelength sensing accuracy and multiple target separations through a communication setting, which is inherently bistatic and noncoherent, (e.g. the transmitter and receiver are positioned differently and not in sync). To address this challenge, we develop algorithms and tools that leverage the spatial diversity of a multiple antenna system or the frequency diversity of a wideband single antenna system to detect and separate close-by targets accurately. The former utilizes multiple antennas at the receiver to observe different signal mixtures from the multiple targets and separate them via blind source separation technique. The latter exploits the phase differences among multiple paths at different frequencies to create different signal mixtures from the multiple targets, then separates them in the frequency domain by suppressing individual signals sequentially. Both methods allow us to detect and characterize tiny motions such as chest movements due to respiration, which we use to verify a patient’s identity during at-home sleep apnea test or establish a trusted link between two devices observing the same patient.

Wireless Physical Layer Security. NextG wireless network systems will connect billions of heterogeneous Internet of Things (IoT) devices to billions of people and enable machine-to-machine communications. The proliferation of low-cost devices with wireless connectivity presents grave security challenges as massive and automatic key exchanges are impractical within a heterogeneous network. Therefore, great interest has been drawn to the physical layer of the communication stack, seeking to leverage the uniqueness of the wireless channel, location, motion, time, trajectory, velocity, and physiological/biokinetic signals to enhance security. However, physical properties often contain insufficient entropy (randomness) and are prone to prediction-based attacks once the adversary learns the physical property characteristics.

To overcome the limitations, we design protocols that utilize reconfigurable devices, such as pattern-reconfigurable antennas, to randomize a slow-fading, quasi-stationary wireless channel and create an artificial fast-changing channel that is difficult to predict. We develop compressed sensing algorithms for the transmitter to calculate the randomizing effect of unused antenna patterns and pre-equalize the channel for the intended receiver. As a result, the main channel state appears stable in the eye of the intended receiver but randomly changes from the attacker’s perspective, which we exploit for secure communication or physical layer key generation (Fig. 4, col 3).

Wireless Communication with Intelligent Reconfigurable Surfaces. With increasing numbers of mobile devices per user, the NextG wireless networks will serve more devices, operate with low latency and less power, and provide functionalities beyond communication. However, the spectral capacity of today’s sub-6GHz wireless networks is already operating close to Shannon capacity limits. To meet the rapidly growing needs of modern devices, high-frequency bands are being considered to host the future wireless networks as they provide wider bandwidth, directional transmission, and servess more users (while considering limited signal coverage and reduced link reliability).

To address the coverage issue at the mmWave band, we develop cost-efficient coverage extension devices, known as intelligent reflecting surfaces (IRSs), which consist of large numbers of low-cost passive/active reflecting electronic elements with reconfigurable parameters. Once deployed, they can capture signal energy proportional to their areas and re-radiate in the shape of beams towards other directions determined by the processors or base stations, allowing NLoS or obstacle-penetrating communication. We also increase the channel diversity between nearby users in outdoor and indoor settings, which reduces interference and increases the number of servable users per area (user capacity).

Federated Learning. Recent advances in electronics with enhanced capabilities of data collection and processing, along with NextG wireless systems with enhanced connectivities, creates a new paradigm in learning, namely federated learning. In federated learning, machine learning models are trained on the edge devices, instead in central servers, for reduced delay and improved privacy. We will develop distributed training algorithms for federated learning with provable performance and reduced communication overhead.