TUTORIALS

T1: Reconfigurable Intelligent Surfaces for Future Wireless Communications (1 day)

Organiser: Alessio Zappone (University of Cassino and Southern Lazio), Marco Di Renzo (CNRS & Université Paris Saclay), Dinh-Thuy Phan-Huy (Orange Labs Paris), Merouane Debbah (Huawei France R&D)

Abstract: As 5G networks take their final form, connectivity demands continue to increase exponentially and new services pose more constraints on the performance that end-users expect. A recent technological breakthrough that holds the potential to meet these demands is that of reconfigurable intelligent surfaces. We believe that a tutorial on the principles and latest approaches of reconfigurable intelligent surfaces for beyond 5G wireless communications will be of great value for both academics and industry practitioners.

T2: A Software Defined Radio Teaching and Research Platform using the RFSoC 2×2 Board (1 day)

Organisers: Robert Stewart (University of Strathclyde), Patrick Lysaght (Xilinx Inc.), Louise Crockett (University of Strathclyde), Cathal McCabe (Xilinx Inc.), David Northcote (University of Strathclyde).

Abstract: In this tutorial we present a single board, fully integrated Software Defined Radio platform for teaching, research and design.  Working live on the tutorial we will feature the Xilinx University Program (XUP) RFSoC 2×2 Board which features 4GHz sampling rate RF ADCs and RF DACs, and an ARM based processing system and FPGA programmable logic facility.  The RFSoC 2×2 Board uses the PYNQ open-source framework and an easy to use browser based system interface exploits features of Lunix, Python and Jupyter notebooks.  In the tutorial attendees will learn how to take direct off the air signals for TV, radio, mobile, wireless, and so on, and downconvert, channelisatize and then investigate the received signal features, including spectral analysis, modulation schemes and other features.  A key feature of the RFSoC DACs and ADCs is their ability to receive and transmit in higher order Nyquist bands, and the tutorial will therefore feature architectures to directly receive and transmit in the 2nd order Nyquist bands (2GHz to 4GHz on the RFSoC 2×2.

T3: AI with Matlab (1/2 day, morning)

Organiser: Stefano Olivieri (MathWorks), Julia Hoerner (MathWorks)

Abstract: AI is everywhere. It is not just powering applications like smart assistants, machine translation, and automated driving, it is also giving engineers and scientists a set of techniques for tackling common tasks in new ways. In this hands-on session, you will get an overview of real-world applications, an introduction of the different techniques (machine learning and deep learning), and practical experience of applying machine and deep learning in MATLAB on different data sets.

T4: Cross-Layer Inference in IoT – From Methods to Experimental Validation (1/2 day, morning)

Organiser: Indrakshi Dey (CONNECT, Trinity College Dublin and Maynooth University), Nicola Marchetti (CONNECT, Trinity College Dublin)

Abstract: The digital transformation is pervading almost every aspect of human life, ranging from healthcare to industry, from entertainment to communications and security. In this respect, the Internet-of-Things (IoT) paradigm plays a crucial role, with a multitude of networked devices interacting with the physical world and providing services through data collection, communication, processing and control. However, several applications require sophisticated design of tailored techniques to perform a wise capitalization of the extracted information with respect to the utilized resources. Energy-efficient IoT design, for example, enables long-life operations with a significant OpEx cut, due to reduced replacement and maintenance costs. Accordingly, this tutorial adopts a statistical signal processing perspective and focuses on the distributed version of the binary-hypothesis test which supports several energy-efficient IoT practical applications concerning the robust detection of a phenomenon of interest (e.g.  environmental hazard, oil/gas leakage, forest fire). The reference scenario is a wireless sensor network and a fusion center with multiple antennas collecting and processing the information. The presence of multiple antennas at both transmit and receive sides resembles a multiple-input multiple-output (MIMO) system and allows for utilization of array processing techniques providing spectral efficiency, fading mitigation, and low-energy sensor adoption. The problem is referred to as MIMO decision fusion. The objective of this tutorial is to cover both design and evaluation (both simulated and experimental) of fusion approaches for this futuristic IoT setup. 

T5: Wireless Powered Communications: A New Communication Paradigm (1/2 day, morning)

Organiser: Ionnis Krikidis (University of Cyprus), Constantinos Psomas (University of Cyprus).

Abstract: Conventional energy-constrained wireless systems such as sensor networks are powered by batteries and have limited lifetime. Wireless power transfer (WPT) is a promising technology for energy sustainable networks, where terminals can harvest energy from dedicated electromagnetic radiation through appropriate electronic circuits. The integration of WPT technology into communication networks introduces a fundamental co-existence of information and energy flows; radio-frequency signals are used in order to convey information and/or energy. The efficient management of these two flows through sophisticated networking protocols, signal processing/communication techniques and network architectures, gives rise to a new communication paradigm called wireless powered communications (WPC). In this tutorial, we discuss the principles of WPC and we highlight its main network architectures as well as the fundamental trade-off between information and energy transfer. Several examples, which deal with the integration of WPC in modern communication systems, are presented. Specifically, we study some fundamental network structures such as the MIMO broadcast channel, the interference channel, the relay channel, the multiple-access channel, and ad-hoc networks. The integration of WPC in 5G and beyond is analyzed and discussed through the use of tools from stochastic geometry. Future research directions and challenges are also pointed out.

T6: DeepFake generation and detection (1/2 day, afternoon)

Organiser: Luisa Verdoliva (University Federico II of Naples), Matthias Niessner (Technical University of Munich)

Abstract: With the availability of powerful and easy-to-use media editing tools, falsifying images and videos has become widespread in the last few years. Coupled with ubiquitous social networks, this allows for the viral dissemination of fake news. This raises huge concerns on multimedia security. This scenario became even worse with the advent of deep learning. New, sophisticated methods have been proposed to accomplish manipulations that were previously unthinkable (e.g., deepfake). This tutorial will present the most relevant methods for generation and detection of manipulated media. For generation, the main techniques based on deep learning will be presented, with focus on those based on both graphics and neural network-based methods such as generative adversarial networks or cutting-edge neural rendering techniques. Both images and videos will be considered, but also the combination of multiple modalities including audio, and text associated with the underlying imagery. For detection, the most reliable deep learning-based approaches will be presented, with focus on those that enable domain generalization. Results will be presented on challenging datasets and realistic scenarios, such as the spreading of manipulated images and videos over social networks. In addition, the robustness of such methods to adversarial attacks will be analyzed.

T7: Multichannel Audio Processing from a Model-based Perspective (1/2 day, afternoon)

Organiser: Jesper Rindom Jensen (Aalborg University), Mads Græsbøll Christensen (Aalborg University), Liming Shi (Aalborg University).

Abstract: The concept of using models of audio and speech in different signal processing applications has existed for decades. Such model-based approaches have many merits, including making any assumptions explicit. This makes it possible to understand their capabilities, limitations, and performance in contrast to the recently celebrated data-driven approaches based on, e.g., deep learning. Despite the many merits, sceptics are often arguing that it is too difficult to estimate the model parameters in practice, and that the model are inaccurate in describing real, natural signals. However, with the recent advances that has been made in speech and audio modeling, it is now clear that they can be robust against challenging phenomena such as non-stationarity and reverberation, and that they can be used to derive both statistically and computationally efficient estimators. Moreover, even if the robust models are still not perfectly matching real signals, it has been demonstrated that the parametric approach outperform the traditional non-parametric one (e.g., correlation-based methods for pitch and direction-of-arrival estimation) in many cases and applications. In this tutorial, we will mainly focus on advances in audio and speech models for acoustic array processing. More specifically, we shed light on the advantages and merits of the model-based approach in different array processing applications such as multichannel enhancement, acoustic source localization, pitch estimation, beamforming, hearing-aids, creation of sound zones, distributed processing and clustering of audio and speech, and robot audition.

T8: When Optical and Wireless Networks Converge to Enable Multi-Service Communications: From Theory to Practice (1/2 day, afternoon)

Organiser: Arman Farhang (Trinity College Dublin), Lei Zhang (University of Glasgow) ,  Colm Browning (Dublin City University)

Abstract: Customized physical layer design, efficient resource allocation, and convergence with the optical world, as enablers for the future high-speed multi-service wireless communication systems, are the main motivations for this tutorial. This tutorial will provide an overview of the advances in waveform design of wireless networks during the last decade, while covering the emerging field of optical-wireless convergence. According to the recent standardization activities, orthogonal frequency division multiplexing (OFDM) with mixed symbol/subcarrier spacing (i.e., mixed numerologies) will be deployed to serve a multi-service ecosystem in 5G networks. Accordingly, a wide set of use-cases are accommodated by an infrastructure at different frequency bands or time slots. However, cohabitation of the individually optimized services in one system leads to technical challenges to both physical and multiple access control layers of the overall system, such as optimal waveform design, inter- numerology-interference analysis and cancelation, resource allocation, etc. These issues are further exacerbated as the shift toward network resource centralization necessitates increasing physical interaction between 5G networks and high-bandwidth optical distribution systems. This tutorial will cover all of these aspects while discussing the emerging candidate wireless and optical technologies that can address the requirements of networks within the 5G era, and beyond.

T9: Massive and Ultramassive MIMO Hybrid Beamforming for Radar and Communications: Optimization to Deep Learning (1/2 day, afternoon)

Organiser: Kumae Vijay Mishra (United States Army Research Laboratory), Ahmet Elbir (Koç University Turkey)

Abstract: The millimeter-wave (mm-Wave) massive MIMO communications employ hybrid analog-digital beamforming architectures to reduce the cost-power-size-hardware overheads. Lately, there is also a gradual push to move from the millimeter-wave (mmWave) to Terahertz (THz) frequencies for short-range communications and radar applications in order to exploit very wide THz bandwidths. At THz, ultramassive MIMO is an enabling technology to exploit even wider bandwidth while employing thousands of antennas. The design of the hybrid beamforming techniques requires the solution to difficult nonconvex optimization problems that involve a common performance metric, i.e. spectral efficiency or energy efficiency as a cost function and a number of constraints related to the employed communication regime and the adopted architecture of the hybrid system(s). There is no standard methodology for solving such problems and usually, the derivation of an efficient solution is a very challenging task. Since optimization-based approaches suffer from high computational complexity and their performance strongly relies on the perfect channel condition, we introduce deep learning (DL) techniques that provide robust performance while designing a hybrid beamformer. The DL methods have gained much interest recently for solving many challenging problems in processing speech, image, biomedical, and remote sensing signals. These methods offer advantages such as low computational complexity and the ability to extrapolate new features from a limited set of features contained in a training set. In this tutorial, the audience will learn about applying DL to various aspects of hybrid beamforming including channel estimation, antenna selection, wideband beamforming, knowledge transfer across various geometries, and spatial modulation.