Research

While utilizing programmable metasurfaces as reflecting devices (i.e. RISs) in the propagation environment has gained tremendous attention in the last few years, the integration of these surfaces into the transceiver equipment is also gaining attention now as an efficient implementation of ultra-massive MIMO and holographic MIMO (HMIMO) communication systems. The conventional implementations of these  systems integrate massive number of active elements at the transceiver and perform signal processing in the digital domain which imposes large hardware and computational costs. Programmable metasurfaces offer low-cost energy-efficient alternative implementations for these systems, and their beam tailoring capabilities can be exploited to realize signal processing in the native analog/EM wave domain. However the number of meta-atoms that can be integrated in a single-layer metasurface is limited, which restricts its ability to implement beamforming effectively in the wave domain. To address this limitation, a stacked  intelligent metasurface (SIM)-enabled base station (BS) architecture shown above has emerged very recently, in which multiple metasurface layers are integrated with the conventional radio transceiver that employs a small number of active antennas.  The multiple stacked metasurface layers offer enhanced signal processing capabilities in the wave domain as  compared to its single-layer counterpart, and can potentially implement digital signal processing completely in the wave domain within the SIM layers at the speed of light. A consequence of such an implementation in a multi-user setup where the SIM response is optimized to perform precoding/combining directly in wave domain is that the number of radio frequency (RF) chains matches the number of users, and since each antenna only handles a single data stream so, we could use low-resolution ADCs/DACs.  

In this context, this research aims to identify transceiver operations and capabilities that can be fully or partially implemented within the SIM layers, such as beamforming, precoding/combining, data detection, and channel estimation, in order to shift the computational burden from the baseband domain to the RF domain, and enable fast, efficient signal processing. The role of SIM to realize other desired communication goals such as index modulation, improved physical layer security, integrated sensing and communication, localization will also be explored. The research focus is on optimization problems that match the performance of wave-based signal processing with digital signal processing solutions, and find appropriate number of metasurface layers for different performance objectives taking into account practical correlation and mutual coupling constraints.  Further, the multilayer SIM architecture is similar to a neural network, which could be further developed to perform more complex deep learning tasks in real-time physically within the SIM layers based on feedback from the environment and/or users. Developing accurate propagation models for inter-layer transmission coefficients that are both physically consistent and mathematically tractable constitutes a fundamental research issue in this area and will be investigated in this research. In addition to studying SIM-enabled implementations for ultra-massive and holographic MIMO communication systems, this line of research aims to characterize the performance limits of these systems in different scenarios involving near-field/far-field communication with/without mutual coupling considering more conventional implementations, and  study the super-directivity gain that dense HMIMO surfaces can yield. 

Reconfigurable intelligent surfaces (RIS) constitute a large number of sub-wavelength passive reflecting elements arranged compactly together with low-cost low-power active components like diodes to realize a programmable structure. By controlling the reflection coefficients of the elements, the RIS can reflect incoming electromagnetic (EM) waves in desired directions to realize different communication, localization and sensing objectives. RISs are expected to play a pivotal role in next generation communication systems and networks due to the unprecedentedly high spectral and energy efficiency gains they offer. This research aims to develop new techniques for achieving the maximum gain from RIS in various applications in wireless communications, and to bring RIS-assisted communication models and techniques closer to practice. Practical and low-overhead channel estimation schemes will continue to be developed to obtain the channel state information (CSI) of RIS-assisted links, and the end-to-end performance of RIS-assisted systems will be analyzed using tools from information theory and random matrix theory. The research aims to optimize the derived performance metrics by designing precoding and power allocation at the base station (BS), and reflect beamforming at the RIS  taking into account practical challenges arising due to CSI inaccuracy, and finite phase resolution of RIS elements. Future wireless networks are expected to transform to a unified communication, sensing, and computing platform with embedded intelligence and programmability. This research will therefore aim to develop integrated sensing, communication and localization schemes with RIS configurations. The goal will be to leverage the ability of RISs to control, sense, and optimize the wireless propagation environment, and realize multi-functional systems that integrate communication and sensing, localization and sensing, and communication and computing. The wireless positioning and sensing of humans, objects, robots, and other mobile devices with RIS will be explored in conventional as well as extreme environments. The research will also focus on the integration of RISs with other emerging 6G paradigms, and identifying and studying good use-cases for RIS as relays and sensors in the these paradigms.

Experimentation with RIS Panels

While the prospects of using RISs to aid different wireless communication, localization and sensing systems is receiving much attention now, most papers conduct theoretical studies based on elementary models, while the prototyping of RISs, as well as experimentation with such prototypes to demonstrate the gains of this technology in practical settings is scarce. To bridge this gap, we will fabricate RIS panels and use them along with vector network analyzers, USRPs, and RF front end equipment for different experimental work. Specifically, this line of research will develop datasets of received signal strength and channel measurements for different RIS-assisted environments (indoor/outdoor, specular/non-specular reflection), and  build experimental proof-of-concepts to demonstrate the potential of RISs in enabling communication to and localizing and detecting users that cannot be seen by the transmitter.

Integration of non-terrestrial network components, such as satellites, unmanned aerial vehicles (UAVs) and high-altitude platforms (HAPs), in existing terrestrial architectures involving the BSs, user equipment and core network, can play a significant role in the future communication scenarios. It is anticipated that 6G and beyond systems will rely more and more on non-terrestrial (NT) components to offer their services globally. This is due to their unique capabilities in extending coverage in areas where a terrestrial infrastructure is impossible or cost-inefficient to reach, as well as their complementary role in offloading an important part of the traffic especially in highly congested areas. Nevertheless, the technical peculiarities coming from the presence of the NT channel, such as larger propagation delays and losses, extra Doppler effects, space weather effects, which are quite different from those of a terrestrial one, bring various challenges to be solved and raise the need for novel solutions for integrated space-air-ground networks.

In this context, this line of research will investigate different integrated space-air-ground network architectures, such as where the NT platform acts a user, the NT platform acts as a relay, or the NT platform acts as the BS. It will also explore mixed architectures where for example: a satellite with BS functionalities may serve the on-ground users through the help of LEO/UAV platforms acting as relays in the access link, or may directly serve the UAVs which will act as aerial users. For these different architectures, the focus will be on problems related to adaptive beamforming and beam management at different nodes, UAV and satellite placement and trajectory design, NT network support of localization services, resource management and optimization in these integrated systems, and machine learning and AI for integrating multi-layered terrestrial and non-terrestrial networks.

There is a growing interest to complement the traditional model-driven design approaches with data-driven machine learning (ML) based solutions in the context of wireless communications research. This research will study the applications of ML approaches to solve complex emerging problems in 6G frameworks such as channel estimation, beamforming design, integrated sensing and communication, localization and detection, radio mapping and so on, without explicit mathematical modeling and analysis. We will use ML tools to design ‘non-linear’ massive MIMO systems that result from the use of one-bit ADCs and DACs in the RF chains to improve the energy efficiency of massive MIMO technology. These systems are hard to study analytically due to the non-linearity of the end-to-end channel transfer function. We will therefore use auto-encoder based deep learning approaches to develop precoders, receivers and constellation designs that are nearly-optimal for these non-linear systems and evaluate the resulting system performance.  This line of research will also focus on the use of ML tools to design localization, sensing and detection schemes to meet different localization and sensing performance objectives. For a given environment, radio features of a mobile device can uniquely determine its location, which makes it possible to perform localization based on radio features. However, the relationship between the mobile location and the corresponding radio features is usually very involved. A deep learning model can be used to map the radio features into a specific location and is something this line of research will investigate. To train the deep learning model, a huge training data on radio features corresponding to different locations is required, which is sometimes challenging. To overcome the overhead associated with collecting all data on radio features from all mobile devices in a certain area and then using it to train the deep learning model for localization, we will also consider federated learning approaches to train deep learning models for radio features based localization.