Our Projects

Object tracking using cameras is a hot research topic with many practical uses, from video surveillance and self-driving cars to analyzing crowd behavior and understanding traffic scenes. The idea is to use one or multiple cameras to follow and identify the location of objects, like people or cars, across several video frames. While this sounds straightforward, it's quite challenging due to factors such as changes in lighting, camera angles, and objects blocking each other.

In recent times, the use of multiple cameras for surveillance has grown due to the availability of affordable, high-quality cameras and powerful computers. Multi-camera systems can offer more comprehensive tracking compared to a single camera, but they also bring additional challenges. For example, ensuring all cameras are in sync, dealing with objects that get blocked from view, and handling changes in how an object looks from different angles.

Yuqiang's project aims to tackle these challenges by enhancing existing methods and introducing a new framework based on machine learning. The goal is to make tracking objects across multiple cameras more accurate and dependable, ultimately contributing to the betterment of real-world applications such as smarter city management and improved traffic flow.

Hydrogen has emerged as an alternative to traditional fossil fuels for powering internal combustion engines as it is carbon neutral, can be produced in a sustainable way and has high energy density. As such, hydrogen engines offer a direct replacement of fossil fueled ​engines with additional benefits coming from the intrinsic properties of hydrogen - these include the ability to operate at ultra-lean conditions with high efficiency and low emissions. However, one of the key factors to unlocking further efficiency gains while retaining the safe and reliable operation of the engine remains the understanding of the autoignition process of hydrogen-air mixtures and how the autoignition is affected by the engine's intake air composition.

Decarbonisation of transport is a major challenge facing nations worldwide. There have been targets put in place by governments to limit the use and sale of petrol-powered automotive vehicles by 2035, meaning that it is becoming increasingly important to develop and improve on existing technologies to power automotives without the use of fossil fuels.

One such method to power automotives is using fuel cells to generate electricity from hydrogen or hydrogen rich molecules. Current fuel cells operate at around 40-60% efficiency in conversion from fuel to electricity which is not enough to one day act as a replacement for petrol in automotives.

There are various methods which can aid in improving the efficiency of fuel cells and one of these is the focus of this research. By increasing the surface area of the commonly used platinum catalyst layer in the fuel cell the rate of the hydrogen oxidation and oxygen reduction reactions that occur in fuel cells can be increased, therefore improving its efficiency. This method utilises growing platinum nanostructures inside a lipid template to create the high surface area structure. This also has the benefit of using less platinum and reducing the cost the material used in the catalyst layer. 

The research will build upon previous PhD students work in which the lipid phytantriol was used as a template to create the nanostructured platinum and will focus in the optimisation of the structure of the platinum in order to produce the highest surface area.

Data from the World Health Organisation (WHO) shows approximately 1.3 million people die annually from road crashes, which are identified as the leading cause of death for children and young adults. In the UK, there were 24,530 people killed or seriously injured in 2021 according to the estimation of the Department for Transport (DfT). Besides concerns on the road safety aspect, road traffic crashes cost most countries 3% of their gross domestic product, leading to considerable financial loss to individuals, their families, and the entire nation.

Meanwhile, various studies prove that human error was the sole factor in more than 50% of road accidents, and was a contributing factor in over 90%. Commonly seen human errors such as drowsy driving, distracted driving, and chemical impairment caused by alcohol or drugs form part of today’s road traffic system, threatening everyone’s life safety.

However, the current development in autonomous driving can’t fully mitigate this issue since the takeover by a human driver is still needed before the SAE level 5 is reached, which is decades away. Propelled by societal pressure and legislation, Driver Monitoring System (DMS) was introduced by car manufacturers to tackle this long-existing problem, combining driver behaviour obtained from a camera and driving behaviour from the vehicle itself to determine the driver’s state. Despite the effectiveness of existing commercial systems, the lack of direct measurement remains a challenge to further improve the accuracy. On the other hand, the feasibility of extracting physiological information such as vital signs based on non-contact approaches in the lab environment has been proven.

Therefore, the focus of Gengqian's project is the development of a novel non-contact driver monitoring system for attentiveness detection via radar, camera, or ultrasonic sensors. Firstly, physiological information is obtained by signal processing and then compared with the ground truth from body-attached sensors to develop a robust non-contact vital sign monitoring system. On this basis, extracted features such as heart rate, respiratory rate, skin temperature, and body movements are combined with observations from real-world driving experiments and brain activity measured by EEG to develop a new model of driver attentiveness. For example, a reduction in heart rate, respiratory rate, or blink rate could be good indicators of low attentiveness.

The decarbonisation of the UK’s nearly 5.5m small and medium-sized enterprises (SMEs), who contribute approximately 1/3 of national emissions, is essential for achieving the national 2050 net zero commitment, as well as for meeting various regional decarbonisation targets. For SMEs, taking steps to reduce operational emissions present opportunities, such as reduced energy bills, generation of additional revenue, attraction of environmentally conscious customers, and futureproofing against regulatory and market pressures. Despite this, SME decarbonisation progress has so far been slow, hindered by key barriers such as upfront cost, perceived unavailability of suitable technologies, lack of information and expertise, and uncertainty surrounding financing, energy markets and changing regulations.

Oliver’s PhD will develop a decision-making tool to help SMEs identify and prioritise investments in energy and transport decarbonisation measures. The tool will optimise both the choice and timing of investments under uncertain future conditions, such as fluctuating energy prices and evolving regulation. Using Bath City Football Club as a case study, the research will model the financial and emissions impacts of feasible decarbonisation options, including use of EVs as distributed energy resources (DERs), and will also explore opportunities for SMEs to coordinate with local network operators, reducing risks and costs for both. Ultimately, this work aims to accelerate SME decarbonisation and integrate their actions into wider energy system planning, supporting regional and national net zero goals.

The automotive industry is under increasing pressure to reduce its environmental impact and contribute to global net-zero goals. Alongside technological changes, such as the shift towards electric vehicles, there is growing attention to Artificial Intelligence (AI) as a potential solution to help reduce emissions. AI is being promoted as a tool for making transport more efficient, reducing waste in production, and supporting greener mobility options. However, while many promises are made about AI’s contribution to net-zero, much less is known about how these claims are communicated to the public and stakeholders, and how this discourse has evolved over time.

This PhD project explores how the automotive industry talks about AI in the context of net-zero and investigates how AI is framed as part of the climate transition. It focuses on how firms, media, and scientific actors construct meaning around AI and sustainability, how they use communication to build legitimacy, and how these narratives shift across time and between actor groups.

The project consists of three academic papers. The first paper explores how automotive firms frame AI within their sustainability narratives, comparing framing strategies across different types of firms (e.g., legacy vs. new entrants, large vs. small, regional differences). The second paper focuses on how media outlets amplify, challenge, or reshape these narratives, looking at differences between general media, industry media, and corporate-affiliated outlets. The third paper looks at the role of scholars, investigating how academic discourse aligns with or diverges from industry and media narratives, and how scientific framing of AI in sustainability contexts evolves over time.

The thesis will conclude with a synthesis chapter that integrates findings across all papers, providing a cross-actor reflection on framing dynamics. It will explore how different actors influence each other’s language over time, which frames become dominant or marginalized, and what this means for the future direction of AI in net-zero transitions. This integrative analysis will offer broader insights into discursive power relations. It will highlight implications for responsible innovation and sustainability governance in the automotive sector.

Berat will investigate a promising solution to increasing emission concerns in air transport is electric urban air mobility (UAM) vehicles, which can provide new mobility models offering zero carbon footprint and be a mainstream mode of air transport in the next decades. However, UAMs cannot offer a ride experience as smooth as that of passenger jets since they will experience complex flight regimes analogous to those of helicopters, such as hover to forward flight transition and high-speed cruise, resulting dynamical loads transferred to the passenger cabin.

The dynamical loads are felt by the occupants as oscillations and vibrations, which can lead to an uncomfortable ride. This is deemed to be critical in UAMs since they will be dominantly used in human transport and their market acceptance depends on the passenger comfort perception. Therefore, UAMs are expected to face various comfort related problems, which needs to be addressed using novel tools and methods.

​Lithium-ion batteries (LIB) have become core technology for energy storage and electric vehicle applications due to key advantages like high energy density, long cycle life, and low self-discharge rates. However, they inevitably degrade over time due to irreversible physical and chemical changes, ultimately leading to the end of their usable life. An accurate and comprehensive degradation model would unlock new opportunities for battery use and optimization.

This research will apply a physics-based electrochemical-thermal battery degradation model coupled a data-driven neural network model to predict the State of Health (SOH), which refers to the battery’s current capacity or performance relative to its original condition, and the Remaining Useful Life (RUL), which estimates the amount of time the battery can continue to function effectively before it reaches its end of life.

​The recent drive to decarbonise transport is leading to a rapid increase in the number of electric vehicles (EVs). High concentrations of EVs could put excess strain on certain parts of the electrical distribution network, thus requiring expensive grid upgrades. In addition, new public charging infrastructure will need to be installed to meet future demand. However, EV charging demand can be flexible, allowing for the shifting of charging time and power to better align with times of high renewable generation and to perform grid services such as load shifting and frequency response.

Understanding where, when and how EVs charge as well as quantifying the flexibility of that demand is imperative in informing the future infrastructure rollout strategy of distribution network operators and local councils to enable cheaper and faster integration of EVs and renewable generation into the grid, thus accelerating the decarbonisation of the automotive and power sectors.

Isaac's PhD research aims to develop high-fidelity spatiotemporal models that leverage data-driven, Bayesian, and agent-based techniques to forecast EV adoption, simulate EV charging behaviour, and quantify the potential impacts of EV charging.

These models could be used to identify potential vulnerabilities in the distribution network, optimal locations for new public charging infrastructure, and provide estimates for the financial and carbon savings that smart charging can offer.