Aligned Students

This project aims to detect stress through the application of self-learning models, focusing on Electrodermal Activity (EDA) analysis alongside other physiological signals. This research explores the efficacy and the application of various self-supervised learning algorithms, such as Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs), in creating stress detection models for individual physiological profiles.

In this project, I will work with a quadruped robot platform and explore how to develop a cognitive model for the robot. I will research existing cognitive models and choose one that is suitable for the mobile quadruped robot, then redesigning and implementing the chosen cognitive model into the quadruped robot. The purpose of applying the cognitive model into the quadruped robot is to enable it to cooperate with humans in performing different tasks autonomously, which would improve human-robot interactions.

Many governments have imposed legislation regulating user data handling, such as the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) etc. Ensuring organisations comply with these regulations is necessary. Formal methods, which are techniques used to model systems mathematically, provide strong guarantees to prove the organisations’ adherence to privacy regulations. My project focuses on defining a framework based on Bigraphical Reactive Systems (BRSs) to prove the privacy regulations within the systems. BRSs are a universal model of computation used to model systems’ behaviours that change their connectivity and locality during the system’s evolution. BRSs have user-specified reaction rules that are defined mathematically and diagrammatically.  I focus on using diagrammatic notations to allow developers and privacy experts (privacy lawyers) to explicitly visualise the systems and describe updates by rewriting rules that describe system behaviour. My project focuses on modelling and proving notions of providing consent, withdrawing consent, purpose limitations, the right to access, sharing data with third parties and international data transfers.

My project aims to explore the potential for socially intelligent agents to learn human social intelligence and apply it in face-to-face social scenarios. Specifically, I will train socially intelligent agents to observe and participate in human social interactions from an egocentric perspective, to analyse social information in terms of feature space and time dimensions based on multimodal information such as the user’s sequence of postural features, voice feature sequences, and semantic information, to recognise and forecast the users’ social intentions, attitudes, and behaviours, and to provide appropriate and timely feedback accordingly.

This research seeks to explore how I can design a social robot for public spaces. I seek to do this in three ways by (1) embedding and altering existing software engineering practices in conjunction with all stakeholders, (2) implementing robust dialogue capability, such as accurate ASR and ability to respond to any query posed in some capacity, and (3) making the robot inclusive to all, as public spaces are frequented by a diverse population. I plan to evaluate my success via real-world deployments of my robot systems.

Future wireless communication system (Beyond 5G and 6G) would be expected to provide digital connectivity for intelligent systems, such as robot, vehicles, drones, etc, and support a wide range of future applications in transportation, industry, healthcare, etc. However, those applications require the extreme communication requirements, such as low latency, high reliability, as well as high date rate, which leads to significant challenges for future wireless systems. In my project I address the scalable and sustainable challenges by jointly designing, with my supervisors, control and communication systems. The goal is to provide an effective communication solution for future connected intelligent systems.

In my PhD project, I investigate visual perception and metacognition through behavioural modelling, aiming to develop a comprehensive model that integrates sensory evidence, trial-by-trial variability, and task interactions. To explore the cognitive and neural mechanisms underlying these processes, I further explore the neural representation of visual perception and metacognition by applying information-theoretic methods and reverse correlation to magnetoencephalography (MEG) data.

Alzheimer’s disease (AD) is a neurodegenerative disease that presents critical challenges in diagnosis and treatment. Emerging research indicates that AD-related cortical changes, such as cortical thickness (CTh), can appear up to a decade before cognitive symptoms. Accurately measuring CTh can therefore offer a significant avenue of early AD diagnosis and monitoring of clinical progression. Current methods for measuring CTh, however, are time-consuming and methodologically flawed, limiting their utility in research and clinical settings. This underscores the need for a rapid, accurate and scalable CTh measurement technique suitable for both large datasets and individualised applications. The current PhD project aims to utilise recent advances in deep learning to build a tool that can generate accurate whole brain CTh estimates from MR images in under a minute. Leveraging extensive longitudinal data, we aim to map CTh and clinical trajectories over time for healthy and AD populations, ultimately personalising these trajectories to an individual level. Finally, we will evaluate the diagnostic and prognostic utility of CTh estimates for Alzheimer’s disease.

The 5th and 6th generation cellular networks (5G and 6G) with impressive performance provide great potential to exchange data and skills over wireless networks. The integration of advanced communication systems with robotic technology is leading the development of connected autonomous robotic systems, such as edge-enabled autonomous driving. These advancements hold immense potential in various sectors including Industry 4.0, healthcare, and education, particularly highlighted during the COVID-19 pandemic where autonomous robots have been instrumental in mitigating the socio-economic impacts in emergency scenarios where human presence is limited. In this project, task-oriented design techniques are studied to address the challenges from the perspective of co-designing communication protocols, robotic systems, and inference engines.

The goal of the PhD project is focused on developing AI methods to predict and monitor cognitive impairment trajectories in Alzheimer’s patients using MRI data. Overall, it aims to create predictive models that use minimal initial data, such as one or two MRI scans, to map the progression of cognitive decline. This approach seeks to expand predictive understanding for neurodegenerative diseases, providing precise, proactive healthcare solutions. The project promises potential advancements in early detection and tailored treatment strategies for individuals, potentially improving outcomes for those affected by Alzheimer’s and similar conditions.

My research focuses on developing and evaluating gaze behaviours in virtual agents within virtual reality (VR). Specifically, it explores two primary approaches: (1) using saliency mapping to drive the agent’s gaze towards high-interest areas in real-time and (2) user experience evaluation to study how varying gaze patterns influence user comfort, immersion, and social presence during interactions. The project aims to create more natural and socially aware agents by integrating intelligent gaze behaviours, grounded in visual cues and user feedback, improving how virtual agents engage with users in dynamic, immersive VR environments.

Based on the current trend in gesture research we are compelled to question if the role of gesture in Human-robot social interaction is limited to a supporting role and cannot convey deeper meanings? To overcome the current impasse, I would like to explore the gesture generation problem by drawing inspiration from the emergent behaviour paradigm. The main objective of this research is to enable robots to learn how to communicate efficiently by observing human-human interaction. This objective will be achieved by (1) developing a baseline method for learning the principles of human-human interaction autonomously from visual data and (2) allowing the robot to learn and develop in an online manner during human-robot interaction sessions in the wild.

This research investigates how behavioural biometrics can be leveraged to enhance both usability and security in immersive virtual environments. By utilising rich user data captured by standard VR sensors—such as gaze fixations and saccades from eye-tracking and movement dynamics from Inertial Measurement Units (IMUs)—the project models fundamental social signals naturally used in human interaction. The goal is to develop continuous authentication mechanisms that verify identity implicitly, preserving the social presence and interaction flow essential for collaborative virtual spaces.

Dyslexia affects around 10–15% of the UK population, with an estimated six million individuals undiagnosed. Many students pass through the entire education system without receiving the appropriate support. Even those who are diagnosed sometimes receive only a binary label rather than personalized help, this is due to significant backlogs and high pressure to issue quick diagnoses. This project aims to develop a screening system that identifies specific cognitive breakdowns in reading within twenty minutes. By collecting a large, multimodal dataset and applying advanced deep learning models in real time, I can tailor interventions to each student’s needs, ultimately improving the educational experience for those with dyslexia.

The future direction of collaborative robots is shifting from having predefined rules for interaction between human and robot in structured environments, such as industrial settings, to achieving more flexibility in their actions in unstructured environments, such as households and public places. These futuristic robots, trained on learning-based methods, show promising results in simulation and controlled environments. However, when deploying these robots in the real-world ensuring human physical and cognitive safety raises major concerns. This necessitates the development of methods which provide provable safety constraints on robot motion which makes it cognisant of its proximity to humans while performing a task, thus providing formal guarantees that the robot trained on learning-based methods will not come in harmful contact with human. Additionally, studying the effect of these safety constraints on task performance and the ease of collaboration through user feedback in joint action tasks between human and robot would further help improve robot behaviour.

The objective of my project is to apply signal processing and machine learning techniques on data acquired from wearable sensors, such as PPGs and EDAs; and contribute towards healthcare and monitoring in a workplace setting. The first research question brings us to the analysis of the quality of signals acquired by PPG sensors and how it varies based on the configuration and location of the sensor. The final goal is to design prototypes which acquire and process such signals more effectively than state-of-the-art.

TBC

My research focuses on developing real-time 3D scene reconstruction pipelines for autonomous robotic platforms using cutting-edge machine learning methods. It also aims to implement communication schedulers that optimise overall reconstruction performance while balancing the trade-off between timeliness and fidelity of the final output.

My project focuses on advancing human-robot (AI agent) interaction, communication efficiency, and data handling for autonomous robots, such as autonomous vehicles in a warehouse, through a task-oriented edge computing framework. The project aims to design a novel framework that integrates reinforcement learning, deep learning, and data compression to control an autonomous vehicle, ensuring human safety.

My PhD research focuses on co-speech gesture generation and evaluation for embodied conversational agents. I will develop models that generate natural, context-appropriate hand  gestures based on spoken language signals like audio prosody, timing, and transcripts or semantic cues. A key goal is improving semantic relevance and interaction timing, so gestures support meaning rather than merely matching rhythm. Alongside generation, I will build a rigorous evaluation framework combining objective motion/coordination metrics with human-centered assessment of appropriateness, clarity, and social impact. The outcome will be methods and benchmarks that help agents communicate more effectively in various social situations.