Open Research Exeter - University of Exeter
Not a member yet
41213 research outputs found
Sort by
A probabilistic digital twin of UK en route airspace for training and evaluating AI agents for air traffic control
No full textThis paper presents the first probabilistic Digital Twin of operational en route airspace, developed for the London Area Control Centre. The Digital Twin is intended to support the development and rigorous human-in-the-loop evaluation of AI agents for Air Traffic Control (ATC), providing a virtual representation of real-world airspace that enables safe exploration of higher levels of ATC automation. This paper makes three significant contributions: firstly, we demonstrate how historical and live operational data may be combined with a probabilistic, physics-informed machine learning model of aircraft performance to reproduce real-world traffic scenarios, while accurately reflecting the level of uncertainty inherent in ATC. Secondly, we develop a structured assurance case, following the Trustworthy and Ethical Assurance framework, to provide quantitative evidence for the Digital Twin’s accuracy and fidelity. This is crucial to building trust in this novel technology within this safety-critical domain. Thirdly, we describe how the Digital Twin forms a unified environment for agent testing and evaluation. This includes fast-time execution (up to x200 real-time), a standardised Python-based “gym” interface that supports a range of AI agent designs, and a suite of quantitative metrics for assessing performance. Crucially, the framework facilitates competency-based assessment of AI agents by qualified Air Traffic Control Officers through a Human Machine Interface. We also outline further applications and future extensions of the Digital Twin architecture.</p
‘Like chalk and cheese’: accounts of police use of force from formerly incarcerated people and implications for procedural justice theory
No full textWhile our understanding of the use of force is predominantly based on police accounts, this article explores accounts of those on the receiving end, specifically formerly incarcerated people, who are underrepresented in the use of force and procedural justice literature alike. Using in-depth interviews with nine such participants, we find they are often positive, offering the police ‘credit where credit is due’ and also, at times, excusing descriptions of excessive force. Yet, while offering the police empathy and respect, they describe not consistently receiving it back, citing concerns over excessive force, neutrality, handling of mental health crises, and untrustworthiness, which, for some, was more damaging and significant than the use of force itself. These findings partially evidence a ‘procedural justice effect’, but find important nuances around how procedural justice theory and police legitimacy operate in a marginalized, liminal population. Implications for policy and practice, including participant suggestions, are also discussed.</p
Federated Retrieval-Augmented Generation-Based LLM for Enhanced Cyber Threat Detection in the Internet-of-Energy
No full textThe Internet-of-Energy (IoE) represents a transformative integration of digital technologies and AI-driven analytics with energy infrastructure, creating an intelligent ecosystem that optimizes energy generation, distribution, and consumption across interconnected grids, renewable resources, and smart consumer devices. While enabling unprecedented efficiency, this interconnectivity introduces significant cybersecurity vulnerabilities, as each component presents a potential entry point for adversaries seeking to disrupt critical operations. Large language models (LLMs) have shown immense promise in addressing cybersecurity issues with their powerful natural language understanding, semantic reasoning, and robust knowledge representation capabilities. However, LLMs encounter significant limitations in processing sensitive, distributed data and executing real-time threat detection in IoE environments. In this paper, we propose FeRAG, a Federated Retrieval-Augmented Generation-based LLM system for autonomous log analysis, designed to enhance cyber threat detection performance while significantly increasing detection efficiency and mitigating privacy risks. We evaluate our method using both GPT-3.5-turbo and GPT-4o as LLM models, and our experimental results demonstrate remarkable improvements of FeRAG over other LLM-driven log analysis methods in the precision of cybersecurity threat detection. Our research also explores promising opportunities for expanding LLMs’ cybersecurity capabilities through integration with diverse multi-modal data sources, enabling more comprehensive threat detection across the evolving IoE landscape.</p
Comparison of Model Predictive Control (MPC) algorithms to optimise blood glucose in fully closed loop (FCL) systems
No full textBackground and AimsModel Predictive Control (MPC) is emerging within fully closed loop (FCL) systems to offer a promising advancement, by automating glucose regulation for people with Type 1 Diabetes. This article assesses the clinical effectiveness of FCL systems and explores future optimisations by comparison of recent developed systems.Methods and ResultsEvidence suggests that MPC-based FCL systems outperform hybrid closed-loop (HCL) models using Proportional-Integral-Derivative (PID) control, achieving higher time-in-range (TIR, 74.4% vs. 63.7%, P = 0.020) and better postprandial glucose regulation. However, no system has consistently surpassed the clinical TIR target (>70%), with postprandial hyperglycaemia and insulin absorption delays remaining key challenges. Three recent emerging FCL advancements include nonlinear MPC (NMPC) for dual-hormone systems, integrating glucagon to reduce hypoglycaemia, λ-Policy Iteration (λ-PI), an adaptive reinforcement learning model, and pulse-modulated artificial pancreas (PMCL) systems, which mimic natural insulin secretion. We compare features of these three emerging solutions and propose a novel hybrid model which combines benefits from these algorithms, to improve accuracy.ConclusionWhile these innovations show promise in in-silico models, clinical validation is lacking. Key barriers include glucagon instability, CGM inaccuracies, cost, and patient adherence. Future research must prioritise long-term trials incorporating real-world factors such as exercise and dietary variability. By integrating predictive control, adaptive learning, and dual-hormone regulation, FCL systems could transform diabetes management, bridging the gap between technology and full automation.</p
Antifungal Immune Responses and Inflammation in the Cystic Fibrosis Airways
No full textCystic Fibrosis (CF) is an inherited disease where the dysfunction of a single protein (the Cystic Fibrosis Transmembrane Regulator or CFTR), a chloride transporter found in both epithelial and immune cells, results in persistent infections and inflammation in the airways. It has been shown that Aspergillus fumigatus frequently infects the airways of people with CF and contributes to hyperinflammation and lung function decline.
The studies presented in this thesis aimed to obtain an insight into the airway mycobiome of people with CF, as well as getting a broader understanding of the host-fungal interactions. This was done by assessing the immune and inflammatory responses to a range of fungal pathogens by immune cells and epithelial cells.
My studies into the host-fungal interactions showed that the antifungal immune responses are fungus and fungal morphotype specific, and that the differences caused by the impaired function of the CFTR protein varies between host cell types. Aspergillus hyphae are more potent inducers of ROS than Aspergillus conidia, while Rasamsonia conidia are more potent inducers of ROS than Aspergillus conidia. No differences were observed in the antifungal responses between healthy and CF epithelial cells and neutrophils against Rasamsonia conidia. CF epithelial cells showed increased phagocytosis, but decreased killing, of Candida albicans blastospores. Interestingly, we observed that the previously described increase in Aspergillus-induced ROS production in CF neutrophils was no longer present in patients receiving treatment with the CFTR modulator Kaftrio. This strongly suggests that Kaftrio has an additional immunomodulating effect without compromising antifungal activity.
I explored if FRISA (Fungal Ribosomal Intergenic Spacer Analysis) could be a valuable and easy to perform technique to study the mycobiome in clinical samples. Unfortunately, due to shortcomings inherent with the technique, I was unable to obtain meaningful results.
As the CF antifungal immune responses are fungus and fungal morphotype specific and antifungal responses differ based on cell type and between healthy and CF cells, it is important to understand the composition of the mycobiome
3
within the CF airways to guide treatment. Especially during exacerbations, it is imperative to determine the best interventions to optimize clearance of the causative fungus and controlling the inflammation evoked. My results have shown that Kaftrio reduces the hyperinflammation against Aspergillus seen in the CF airways without effecting fungal clearance. And as a result of the widespread use of Kaftrio, and other CFTR modulators, the perspective of CF associated fungal disease may well be changing substantially.</p
Skateboarding’s sovereign excellence
No full textThis paper considers sovereign attributions to skateboarding by various public intellectuals and scholars characterized as ‘heroic’, ‘aristocratic’, and of a ‘higher style of play’. They argue that such attributions are indicative of a form of internal excellence (areté) that manifests externally in a continuum from criminal vandal to Olympic athlete not unlike similar attributions in Archaic and Ancient Greece as well as the Edo period of Japan. They argue further that skateboarding’s sovereign excellence includes subversive elements that present a ritualized reworking of the meaning and value of the city, tacitly redeeming it from a merely pecuniary role. While toying with sovereignty, its value for understanding excellence within the sport enclave, the authors also propose an epistemology that takes seriously the mythic poetics of skateboarding.</p
Physiology of everyday sleep and physical activity: An exploratory mixed-methods study of multi-sensor wearables for infants and toddlers
No full textSleep and physical activity are vital to the health, development, and well-being of young children. To effectively promote these behaviours at the population level, better tools for objectively quantifying them are needed. This hypothesis-generating mixed-methods study explored the potential usability of two wearable sensors to measure physical activity and sleep in young children over multiple days, drawing on physiological measurements. A longitudinal within-case design was employed, in which families with children aged 4–36 months from the North East of England were recruited through playgroups and social networks. Parents and children tested two wearable devices in a structured play setting and at home over a period of 1 week. Data on sleep, movement, and heart rate were collected using the Bittium Faros 180 heart rate monitor and the NAPPA sleep monitoring system. Usability was assessed through researcher observations and parent feedback using ethnographic methods. Wear time, heart rate variability during naps, and ultradian respiration cycles during sleep were analysed. Seven children participated and completed the study. While parents were initially enthusiastic, usability challenges arose. The heart rate monitor was considered uncomfortable, its large size hindered activity, and electrodes were detached by parents and accidently, leading to significant data loss. The NAPPA was easier to use, discreet, and comfortable, but disrupted sleep routines. Additional challenges related to non-parental caregiving resulted in non-wear and/or data loss. These results indicate that wearable devices for young children hold potential but face significant design challenges for longitudinal home use at scale. Co-creation of child-friendly, practical hardware and software is essential for effective, large-scale health monitoring in young children.</p
The Ocean Equity Index
No full textThe ocean is essential for humanity1,2,3. Yet, inequity in ocean-based activities is widespread and accelerating4,5,6,7,8. Addressing this requires governance approaches that can systematically measure equity and track progress9. Here we present the Ocean Equity Index (OEI)—a framework for assessing and improving equity in ocean initiatives, projects and policies. We apply the index, which scores twelve criteria, to case studies at local, national and global scales. We show that the OEI can generate structured data to support evidence-based decision-making across ocean sectors and scales. As a theoretically robust and widely applicable tool, the OEI can guide the design of more equitable ocean initiatives, projects or policies, ensuring better outcomes for coastal people and marine ecosystems.</p
Doubts about the usefulness of masked priming for tracking lexical consolidation: Only short-term effects of exposure and no help from sleep
No full textQiao and Forster (2013) showed that repeated exposure to novel neighbours (e.g., 'banara') leads to the disappearance of facilitation in masked priming when trained items are used as primesa between-subject result taken as a sign that newly learnt primes had become competitors of their targets (e.g., 'BANANA'). Here, we assess the long-term nature of this effect and the possible involvement of sleep in its emergence, by teaching participants two sets of neighbours (e.g., 'alarchy' for 'anarchy') 12 hr apart, and testing them immediately after learning Set 2 and again after another 12 hr. Crucially, half of the participants (PM group) learn Set 1 at 20:00, while the other half (AM group) learn Set 1 at 08:00. Training does reduce the amount of orthographic facilitation (i.e., 23 vs. 32 ms). However, this effect is short-lived and gone within 12 hr, whether participants sleep or not after learning. Also, at odds with the prime-lexicality account, this effect is driven by unrelated targets becoming faster, rather than related targets becoming slower, as a result of neighbour exposure. In contrast, a Reicher-Wheeler task, pitting the neighbour against its base word (e.g., 'alarchy/anarchy'), shows reliable signs of sleep-associated consolidation: the rate of neighbour-consistent responses (e.g., 'l'), made stronger through exposure, remains unchanged during subsequent wakefulness and is enhanced after sleep. Given that the reduction in masked orthographic facilitation following neighbour exposure is clearly a short-term effect, our results call into question Qiao and Forster's conclusions and the usefulness of masked priming for tracking lexical consolidation.</p
Lexical and Collocational Sophistication in Chinese High School Writing
No full textIn corpus linguistics, the discussion of how language users employ word combinations in their writing has attracted researchers’ attention for many years. This thesis explores how Chinese senior high school students make use of their lexical and collocational knowledge in their academic writing. This thesis applies two corpora to compare passages written by Chinese senior high school students and mainstream Year 11 secondary students in the United Kingdom.
In study 1, I explore how two groups of students use collocations and discuss the relationship between collocation strength measures and Chinese students’ writing scores. As for low frequency word combinations, I explore how Chinese and UK students use low-frequency word combinations and discuss the relationship between low-frequency word sequences and Chinese students' writing scores. In study 2, first, I explore how Chinese students and UK Year 11 students apply their vocabulary knowledge in terms of lexical sophistication in academic writing. In this section, many lexical sophistication indices are employed in seven aspects. In addition, I examine the relationship between lexical sophistication indices and Chinese students’ writing scores. Second, I explore how Year 11 students and Chinese senior high school students make use of vocabulary knowledge in terms of lexical diversity. Moreover, I investigate the relationship between lexical diversity indices and the writing scores of Chinese students.
The data reveal significant differences between Chinese students and United Kingdom students in the strength of association measures and the use of low-frequency word combinations. In addition, the association strength measures correlate positively with Chinese students’ writing scores. Higher composition overall scores are associated with fewer low-frequency word combinations. In terms of lexical sophistication and lexical diversity, there are significant differences between the two groups of students. Several indices of lexical sophistication and lexical diversity could influence Chinese students’ overall writing scores. Generally, Chinese students’ writing is influenced by their English vocabulary, their first language, and guidance from the national English teaching curriculum. Language features revealed in the writing of Chinese senior high school students point to important pedagogical implications for teachers and offer suggestions for future researchers.</p