Current Ongoing Projects:
1. ‘Scalable Hybrid Architecture for Wireless Collaborative Federated Learning (SHAFT)’, February 2023- January 2026 (Funded by EPSRC)
Unlike previous generations of mobile networks, the beyond 5G (B5G) network is envisioned to support edge intelligence, which is to provide both communication and computing capabilities to the proximity of end users. Wireless edge intelligence is particularly important to those crucial use cases of B5G, including smart cities, autonomous driving, wireless healthcare, virtual reality (VR) and augmented reality (AR) gaming, where mobile networks are expected to be equipped with intelligent capabilities for prediction and shaping experiences to individuals. Federated learning (FL) is a key enabling technology for wireless edge intelligence, by performing the model training in a decentralized manner and keeping the data where it is generated. However, a straightforward adaption of FL from computer networks to wireless systems can suffer performance degradation in spectral and implementation efficiency, because of the complex wireless environment with heterogeneous resources and a massive number of devices. The aim of this project is to develop a novel scalable hybrid architecture for wireless FL by efficiently utilising the physical layer dynamics of the mobile communication environments and exploiting sophisticated service-aware and resource-aware collaborative edge learning. The novelty of the project is the development of this novel edge learning architecture, where the fundamental limits of the learning architecture is characterised by advanced mathematical tools, such as graph theory and stochastic learning. In addition, an algorithmic framework for quantifying challenging design trade-offs in the presence of practical constraints by applying sophisticated tools such as compressed sensing and machine learning.
2. ‘Smart Solutions Towards Cellular-Connected Unmanned Aerial Vehicles System (AUTONOMY)’, August 2022- July 2025 (Funded by EPSRC)
Unmanned Aerial Vehicles (UAVs) with low cost and high mobility are recognised as an emerging technology that will lead to significant economic growth and broad societal benefits, which are also essential to tackle the COVID-19 pandemic (e.g., via medical supplies delivery or disinfectants spray). The market growth for the commercial UAVs industry is expected to skyrocket to 45.8 billion dollars in 2025 from 19.3 billion dollars in 2020, at a compound annual growth rate (CAGR) of 15.5% from 2019 to 2025. UAVs are of paramount importance for numerous civilian applications in diverse fields, including aerial inspection, precision agriculture, photography, package delivery, traffic control, search and rescue, and telecommunications. Nevertheless, the above benefits can only be reaped with advanced wireless communication tech-niques, intelligent sensing and networking operations, and joint communication and control designs that can support safe UAV operations, mission-specific rate-demanding pay-load communications, and efficient multi-UAVs cooperation. Conven-tional UAVs relying on the short-range communication technologies (e.g., WiFi) with the low data rate are unfortunately insufficient or even inapplicable to support beyond-visual-line-of-sight (BVLOS) communications with wide-area connectivity. These limitations have motivated academia and industry to explore the use of cellular networks in 5G-and-Beyond in providing ultra-reliable low-latency control, ubiquitous cover-age, and seamless swarm connectivity under complex and highly flexible multi-UAV behaviours in three-dimensions (3D), to unlock the full potential of UAVs. This so-called cellular-connected UAVs (C-UAVs) System creates a radically different and rapidly evolving networking and control environment. This project aims to be the first to develop and implement full network automation and conditional control automation (on condition of high trust) for C-UAVs system in simulator and prototype.
3. ‘Distributed Acoustic Sensor System for Modelling Active Travel’, January 2023- December 2025 (Funded by EPSRC)
In a time where climate change is an imminent threat, Active Travel (AT) has become a priority in the United Kingdom (UK) and a pathway towards sustainable living. AT is defined as making a journey by physically active means, e.g., walking or cycling. In the UK, the transport sector is the highest contributor of emissions with 61% of this contribution caused by private cars and taxis. Replacing motored journeys with AT firstly promises to reduce these emissions. Moreover, AT is a form of exercise that has been shown to improve physical and mental health; hence, reduces the need of medical care and increases happiness and productivity. Interventions to promote AT include ensuring safety of commuters through cycle/pedestrian lanes, safe cycle parking, bike-sharing, cycling training, bike loan schemes, electrically assisted bikes, community/school initiatives, among others. The challenge that authorities face is the lack of insights on which type of intervention would be more effective in different areas. Indeed, the same scheme would result in different AT uptake since the latter depends on predominant trends and road infrastructure in each area. It follows that, in each area, some schemes are likely to be more effective than others.
There is a rising need to model changes in AT trends in relation to different interventions. State-of-the-art research for modelling AT trend mostly relies on video footage which is used to identify and predict the path of pedestrians. There are several drawbacks to such approaches. Firstly, video footage is negatively impacted from adverse weather conditions and lack of light. Secondly, it is cost-inhibitive to realise uninterrupted 360 degrees visibility using video cameras in a built environment. Thirdly, the video footage needs to be high resolution, hence contains private information about people. Such information challenges General Data Protection Regulation (GDPR) whilst is not required for modelling active mobility.
DASMATE aims to develop a new approach for modelling AT trends in an urban environment by leveraging the incipient advances in Distributed Acoustic Sensor (DAS) systems. DAS reuses underground fibre optic cables as distributed strain sensing where the strain is caused by moving objects above ground. Given that the sensors are underground, DAS is not affected by weather nor light. Fibre cables are often readily available and offer a continuous source for sensing along the length of the cable. Moreover, DAS systems offer a GDPR-compliant source of data that does not include private information such as face colour, gender, or clothing. DASMATE in centred on two aspects of AT modelling based on DAS analysis. The first consists of identifying the type of AT (walking, jogging, skateboarding, cycling, etc.) at any time of the day in a monitored area. The second is concerned with predicting the path of active travellers to inform on the possibility of collision with moving vehicles (which may be driver-less). This a pioneering project that aims to establish the first framework for processing DAS data to extract samples representing AT and build a machine learning pipeline to infer knowledge related to both aspects.
This project will be worked together with partners both from the industry and UK authorities such as Fotech and London Borough of Tower Hamlet. The principal investigator (PI) maintains a strong track record in signal processing with professional skills machine learning, and optimization. The industry partner Fotech is leading the smart city application of DAS and has been collaborating with PI for a year on DAS-based vehicle classification and occupancy detection. Moreover, a unique DAS dataset for AT modelling that will enable this project has been collected jointed through this collaboration. The London Borough of Tower Hamlet finds value in this project and has offered to trial the technology outcomes in the borough to measure the efficacy of planned AT schemes.
4. ‘Towards Sustainable ICT: Sparse Ubiquitous Networks based on Reconfigurable Intelligent Surfaces’, August 2022- July 2024 (Funded by EPSRC)
The ever-increasing demand for ubiquitous wireless communication services with high data rate, low latency, and high reliability is considered unsustainable from an environmental perspective. In particular, the emerging high-rate services in the millimetre-wave (mmWave) range require a large number of base stations to cover larger areas or urban environments, which leads to a significant increase in the consumption of energy and resources.
Several ideas exist to improve these networks without the need for additional base stations; in particular, reconfigurable intelligent surfaces (RIS) have recently been proposed as sustainable alternatives for many application scenarios. The main idea is to intelligently reflect otherwise unused and wasted signals back onto a path to the user, to improve overall coverage and throughput. However, such RISs typically require a large number of actively tunable components to dynamically redirect the reflections towards the users. The associated energy consumption prevents sustainability from being ensured in this way.
This project builds on the idea of enhancing the wireless links via smart reconfigurable surfaces, but it adds an important key ingredient to actually achieve sustainability, while providing the needed improvement for wireless service coverage: sparsity. Sparse placement of tunable components on these reconfigurable surfaces, then referred to as sparse reconfigurable intelligent surfaces (SRISs), results in significant energy efficiency improvement. Beyond that, employing sparsity also in the deployment of these SRISs, further improves sustainability, by minimizing both manufacturing costs and used resources as well as overall energy consumption.
Key components of the projects include the development of physics-based models of SRIS and a holistic design methodology based on combining ideas from metasurface design and sparsification, channel modelling via efficiency-enhancing electromagnetic simulation techniques, and optimisation of resource allocation and deployment (via modern convex optimisation and relaxation as well as artificial intelligence (AI)-based methods) to maximize SRIS improvement. Prototypes and measurements will prove the working concepts.
This project will pave the way toward future high-performance wireless networks that are energy-efficient and sustainable by design.
Recently Completed Projects:
5. Communications Signal Processing Based Solutions for Massive Machine-to-Machine Networks (M3NETs) (Funded by EPSRC)
Future wireless networks should have the capability of serving a wide range of personal wireless devices and appliances with stringent end-to-end delay requirements. These appliances will be equipped with the capabilities to sense various real time events and be able to self-configure via network connections, thereby paving the way for many emerging applications including e-health, intelligent transportation and smart cities. The most important enabling technologies for these applications is seamless machine-to-machine (M2M) wireless communications, which is the key to sustaining large-scale massive interconnections between things. The number of M2M devices has been growing exponentially, and is expected to reach up to 50 billion by 2020. This trend in the market growth for both M2M devices and M2M connectivity segments will further accelerate in the future. As such, M2M communications is envisioned as one of the five disruptive technology directions for fifth generation (5G) wireless networks and beyond.
Despite the importance of machine-type communications, there are many critical challenges that need to be addressed in terms of network congestion and overload due to presence of massive M2M devices with heterogeneous traffic patterns, unprecedented level of inter and intra interference among M2M and human-to-human (H2M) communications, complex resource management due to irregular traffic patterns and energy constraints. The focus of this project is on tackling these critical challenges, by advancing aspects of communications signal processing, stochastic geometry, convex optimizations and game theory. In particular, we contribute in terms of characterizing heterogeneous traffic patterns associated with massive M2M communications, development of distributed random access channel protocols, proposal of convex and game theoretic resource allocation methods and design of energy harvesting constraint based cross-layer optimization algorithms and protocols.
6. Reinforcement learning for multi-objective optimization in wireless networks (Funded by Huawei Technologies, Ireland)
There has been significant development in 5G networks at the physical/cross layer. Within this broad aim the project focuses on the problem of maximizing coverage and throughput, while maintaining an acceptable complexity. Due to the complexity in mathematics and the difficulty in constructing physics-based model, solving this problem with explicit optimization instructions is generally intractable. The RL operating setup depends on an inference model which maps the historical network performance to future configurations. This leads to a number of multi-objective optimization problems, where in each case the aim is to learn a policy to operate the network parameters to achieve the desired balance in the objectives.
7. Multi-Cell NOMA (Funded by Huawei Technologies, Shanghai)
Non-orthogonal multiple access (NOMA), which has been recently proposed for the 3rd generation partnership projects long-term evolution advanced (3GPP-LTE-A), constitutes a promising technology of addressing the above-mentioned challenges in 5G networks by accommodating several users within the same orthogonal resource block. By doing so, significant bandwidth efficiency enhancement can be attained over conventional orthogonal multiple access (OMA) techniques. This project concentrates on the design of NOMA scheme under multi-cell environment.
8. Enabling High-Speed Microwave and Millimetre Wave Links (MiMiWaveS) (Funded by EPSRC)
Even though wireless channel impairments greatly impact the bandwidth efficiency of wireless networks, their effects have not been taken into consideration in the recent research carried out in this discipline, especially in the microwave and millimetre-wave bands for fifth generation (5G) cellular. The objective of this project was to improve the bandwidth efficiency of next generation 5G operating in the microwave and millimetre-wave bands through effective transmitter and receiver designs that exploit massive multiple-input multiple-output (MIMO) and heterogeneous small cell deployment, while taking into account the effects of impairments, such as channel estimation error, phase noise, and carrier frequency offset. As a result, this project is not based on any idealistic assumptions regarding the wireless channel, which compared to existing work in this field is unique. The proposed research raised several fundamental design challenges far from trivial, that have their roots in diverse disciplines, including information theory, stochastic control theory, sequential statistics, large system analysis, automated decision making, and pervasive computing.
9. Simultaneously Wireless InFormation and energy Transfer (SWIFT) (Funded by EPSRC)
Information and energy are two fundamental notions in nature with critical impact on all aspects of life. All living and machine entities rely on both information and energy for their existence. Most, if not all, processes in life involve transforming, storing or transferring energy or information in one form or the other. Although these concepts are in harmony in nature, in traditional engineering design, information and energy are handled by two separate systems with limited interaction. In wireless communications, the relationship between information and energy is even more apparent as radio waves that carry information also transfer energy. Indeed, the first use of radio waves was for energy transfer rather than information transmission. However, despite the pioneering work of Tesla, who experimentally demonstrated wireless energy transfer (WET) in the late 19th century, modern wireless communication systems mainly focus on the information content of the radio-frequency (RF) radiation, neglecting the energy transported by the signal. This project was the first interdisciplinary initiative to promote innovation and technology transfer between academia and industry in the UK for one of the most challenging and most important problems in future communication networks: The simultaneous transfer of both energy and information. The aim of this project was to develop a new theoretical framework for the design and operation of next-generation networks with simultaneously wireless information and energy transfer (SWIFT) capabilities. The research efforts are interdisciplinary and bring together researchers with strong and complementary backgrounds in the domain of wireless communications such as electronics/microwave engineering, information theory, game theory, control theory, and communication theory to bridge the gap between theory and practice of future WET-based communication systems.
10. Massive MIMO wireless networks: Theory and methods (Funded by EPSRC)
Maximizing spectral efficiency, which is limited by interference and fading for wireless networks including 4G, is a major issue. Massive MIMO technology has the potential to unlock the issue of spectrum scarcity and to enhance spectrum usage tremendously by enabling simultaneous access of tens or hundreds of terminals in the same time-frequency resource.
In order for massive MIMO technology to attain its utmost potential, it is important that various challenges in terms of channel estimation and acquisition due to pilot contamination, fast spatial-temporal variations in signal power and autonomous resource allocation, in particular in the presence of simultaneous access of a large number of users need to be addressed. The focus of this project was on tackling these fundamental challenges, by advancing aspects of information theory, estimation theory and network optimizations. In particular, we contributed in terms of modelling massive MIMO channels underpinned by heterogeneous correlation structures; performing information theoretic analysis in terms of random matrix theory through shrinkage estimators; robust precoder design for massive MIMO in the presence of channel estimation errors; developing novel channel estimation technique in the presence of severe pilot contamination; and proposing and analysing game theoretic algorithms for autonomous resource allocation and pilot assignments. All the concepts and algorithms developed could be integrated in 5G and Beyond 5G systems.