Doctoral Theses (Mathematical & Computer Sciences)
Permanent URI for this collectionhttps://hdl.handle.net/10399/37
Browse
538 results
Search Results
Item Conjugacy properties in classes of Artin groups and their extensions(Heriot-Watt University, 2024-09) Crowe, Gemma Margaret; Ciobanu Radomirovic, Professor Laura IoanaIn this thesis we study conjugacy properties from a combinatorial, formal language and algorithmic perspective, in classes of Artin groups and their extensions. We study the nature of conjugacy languages for right-angled Artin groups and their extensions. We also solve the twisted conjugacy problem for right-angled Artin groups, with respect to length-preserving automorphisms, and consider various examples, using geometric and algebraic techniques, where extensions of right-angled Artin groups have solvable conjugacy problem. For dihedral Artin groups, we compute conjugacy geodesic representatives, and show that the conjugacy growth series is transcendental for dihedral Artin groups, with respect to some generating set. We also prove regularity of the conjugacy geodesic language, by studying the permutation conjugator length function and falsification by fellow traveller property. Finally, we solve the twisted conjugacy problem for all dihedral Artin groups, which leads to new examples of extensions of dihedral Artin groups with solvable conjugacy problem. Key words: Conjugacy growth, conjugacy languages, twisted conjugacy, right-angled Artin groups, graph products, dihedral Artin groups.Item Enhancing myoelectric prosthetic control with deep learning : continuous DOF estimation and generalization(Mathematical and Computer Sciences, 2025-10) ElMohandes, Hend; Elgayar, Doctor Neamat; Taylor, Professor Nicholas Kenelm; Turcanu, Doctor AdrianItem Investigating optimal strategies for power storage facilities(Heriot-Watt University, 2025-08) O’Brien, Fraser; Johnson, Doctor TimothyThe variation in electricity generation is increasing significantly as the UK transitions to a renewable based energy grid, introducing logistical challenges in balancing the supply and demand of electricity. These challenges can be overcome through the use of power storage facilities, which ensure that electricity can be generated when possible and provided to the grid when needed. Private investment in power storage facilities must be encouraged through studies of their profitability, for which suitable methodologies must be developed, with facilities aiming to identify optimal strategies which maximise profits by considering the values of different actions and acting in a manner which adds the most value to the facility. The value of the facility is dependent upon the excess demand - the difference between the demand and supply - of electricity, with increasing excess demand leading to higher prices of electricity. Whilst the value of charging and discharging are known for each facility and given by payoff functions, the value of waiting has historically required assuming the form of the stochastic process assumed to underlie excess demand data and the estimation of parameter values from which the value of waiting can be calculated. Instead, this thesis employs an empirical approach to directly estimate the value of waiting from historical excess demand data. The methodology is derived using traditional stochastic control techniques requiring the underlying process only to be volatile in some manner and non-explosive, requiring no set form or even the existence of a canonical form which means it is suitable for application on data-sets where the underlying dif fusion is unknown such as with excess demand. With an eye on the continued development of power storage technologies a development of standard ‘bang-bang’ models which only consider complete actions is introduced that allows for the consideration of partial actions, allowing for more accurate modelling of developing power storage technologies with quick reaction times to changes in excess demand levels. Furthermore, variations in facility efficiency and electricity pricing - and the resulting effects on the optimal control strategy - with respect to different factors are considered through their affects on the payoff functions, alongside the potential interaction between factors to provide a methodology to define a complete optimal control strategy.Item Explainable reasoning for remote autonomous agents(Heriot-Watt University, 2025-08) Gavriilidis, Konstantinos; Hastie, Helen Frances; Konstas, Ioannis; Pang, WeiRemote autonomous robots are increasingly deployed for demanding tasks such as underwater exploration and pipeline inspection, providing valuable ecological insights and generating commercial benefits. However, human-in-the-loop applications in this domain face significant challenges, including a lack of direct supervision, bandwidth limitations, and limited technical understanding of the underlying autonomous systems. Ensuring situational awareness and trust is critical for the broader adoption of these technologies. This research project addresses these challenges by developing novel methodologies for transparent and explainable autonomy. The work focuses on two primary objectives: generating explanation content and effectively communicating it through natural language. To achieve the first objective, domain knowledge and robot state fusion are employed, alongside the creation of simplified autonomy models using surrogate techniques. For the second objective, the communication of explanation content is explored using both template-based and language model-based approaches, supporting causal, counterfactual, and contrastive explanations. User preferences for these explanation styles are evaluated, and the effectiveness of model-based explanations is compared to that of template-based alternatives. The findings demonstrate satisfactory performance in approximating autonomy using both surrogate and language models. Moreover, this work identifies the explanation styles that most significantly enhance situational awareness. These results contribute to the advancement of transparent and explainable autonomy, facilitating greater trust and adoption of remote autonomous robots in challenging applications.Item Building and breaking trust : how failures and repairs affect people’s trust in robots(Heriot-Watt University, 2025-07) Nesset, Birthe; Hastie, Professor Helen; Rajendran, Professor Gnanathusharan; Romeo, Doctor MartaWith the increasing integration of robots into social and industrial domains such as healthcare and manufacturing, understanding how trust is built, maintained, and repaired in the face of system failures becomes crucial to ensure the long-term effectiveness and acceptance of autonomous systems. This thesis aims to contribute to the research field of Human-Robot Interactions (HRI) by investigating different aspects of trust in the face of robot failures and repair attempts. To do so, five separate studies were conducted. First, the thesis evaluates different trust measures to ensure that the correct tools are used throughout the thesis and to provide recommendations for other researchers. Then, two studies investigate repeated failures and repair strategies through the implementation of a communication incongruence performed by both a robot and a human actor. Finally, the last two studies examine humans’ attitudes towards artificial “trust” and investigate if trust in HRI can be transactional. The work presented in this thesis consists of both qualitative and quantitative findings, all contributing to a better understanding of how trust in HRI can be impacted by failures. Key findings in this thesis include guidance on trust repair in the face of repeated failures, a trust scale comparison, and reflections and recommendations surrounding artificial “trust” implementation.Item SDMANET : enhancing MANETs performance with hybrid protocols through SDN integration(Heriot-Watt University, 2025-05) Saleh, Rabia; Ibrahim, Doctor Idris Skloul; Georgieva, Doctor LiliaIn the ever-changing landscape of applications, traditional Mobile Ad hoc Networks (MANETs) have to overcome significant hurdles to optimise bandwidth, manage node mobility, and efficiently handle power. These limitations have far-reaching effects on network structure, operational mode, stability, and the performance of upper-layer applications. This thesis proposes an innovative network architecture that aims to address these challenges by integrating Software-Defined Networking (SDN) with MANET. SDN offers a centralised network perspective, enabling improved control and forwarding decision-making. The Software-Defined Mobile Ad hoc Networks (SDMANET) approach represents a sophisticated integration of SDN nodes with several traditional routing protocols, including OLSR, BATMAN, BATMAN-ADV, and GRE. This multifaceted method facilitates efficient IP-based packet forwarding while leveraging the capabilities of the OpenFlow protocol to manage data flows within the SDN framework. By combining these advanced technologies, the SDMANET approach ensures smooth interoperability between modern SDN architectures and traditional legacy nodes. This integration not only enhances network flexibility and performance but also preserves the functionality of existing routing methods, making it an effective solution for evolving network environments. To mitigate the risk of a single point of failure, the experimental setup includes an ONOS cluster instead of a singular centralised controller. Notably, SDMANET outperforms traditional routing protocols in terms of throughput, packet delivery ratio, packet loss, end-to-end delay and routing overhead for both TCP and UDP traffic, demonstrating promising results compared to previous studies (Mishra et al., 2018; El-Garoui et al., 2020; Hakiri et al., 2017). The results demonstrate that SDMANET achieves up to 37% higher throughput, about 30% lower end-to-end delay, and approximately 25% improvement in packet delivery ratio compared to conventional MANET protocols under both TCP and UDP traffic. It also significantly reduces routing overhead in dynamic topologies. These results substantiate SDMANET as a promising solution for next-generation mobile networks, especially in 5G and 6G contexts, where scalability, adaptability, and efficient resource management are critical. This research thus contributes a scalable, performance-oriented SDMANET framework that addresses key gaps in existing architectures.Item Enhancing particle swarm optimization through artificial neural networks(Heriot-Watt University, 2025-04) Benhalem, Amani Mansur L.; Lones, Professor MichaelBackground The integration of Artificial Neural Networks (ANN) within Particle Swarm Optimization (PSO) aims to enhance optimisation algorithms for robotics and signal processing applications. PSO is an evolutionary computation technique inspired by the social behaviours of birds and fish, where particles iteratively adjust their positions based on individual and collective experiences to optimise solutions. Its simplicity and adaptability make it effective, but performance variability due to parameter sensitivity requires further investigation. This study seeks to improve PSO’s efficiency by integrating ANN, addressing limitations and enhancing swarm intelligence in dynamic robotics applications. Methods The study developed a new form of Particle Swarm Optimization (ANN-PSO) and applied it across two domains: continuous optimization and swarm robotics, using both mathematical benchmarks and the Simbad 3D simulator. For continuous optimization, the ANN-PSO's performance was analyzed through rigorous trials across various complex functions to test its capability to locate global optima. In swarm robotics, controlled experiments involved navigating robotic swarms towards targets while avoiding obstacles, comparing ANN-PSO with standard PSO. The experiments assessed metrics such as navigation efficiency, swarm cohesion, and adaptability to environmental changes, focusing on the integration of neural networks within the PSO framework to enhance operational dynamics. Results The dual application of ANN-PSO demonstrated enhanced performance in both continuous optimization and swarm robotics. In continuous optimization, ANN-PSO showed superior capability in navigating complex landscapes, achieving more precise convergence on global optima and displaying robust exploratory behaviors. In robotic swarm experiments, robots programmed with ANN-PSO exhibited superior navigation strategies, characterized by improved decision-making and optimized pathfinding capabilities compared to those using standard PSO, where robots showed less efficient navigation. These results confirm the efficacy of ANN-PSO in managing diverse and dynamic optimization challenges, effectively enhancing both the depth and scope of search processes. Conclusion This study validates the substantial benefits of integrating artificial neural networks with Particle Swarm Optimization, creating a versatile and powerful tool in ANN-PSO that excels in tackling complex problems across continuous optimization and swarm robotics. The adaptability and enhanced learning capabilities imparted by ANN integration not only improve operational efficiency but also broaden the potential applications of PSO in complex and dynamically changing environments. Future research directions include exploring further applications of ANN-PSO in other complex systems, enhancing algorithmic efficiency, and extending its adaptability to more diverse environmental settings.Item Learning to handle miscommunication in multi-modal conversational AI(Heriot-Watt University, 2025-04) Chiyah-Garcia, Javier; Suglia, Assistant Professor Alessandro; Eshghi, Associate Professor ArashIn human communication, we continuously negotiate shared understanding and deal with misunderstandings as they arise to achieve mutual coordination. However, despite the ubiquity and importance of misunderstandings and repairs in dialogue, conversational AI often struggles to process them effectively, limiting their ability to collaborate with humans through natural language. This thesis explores how to develop robust models for processing miscommunications in situated collaborative tasks. We first collect a dialogue corpus to study human-agent coordination in an ambiguous environment, finding that models struggle to resolve referring expressions. To address this shortcoming, we design and train models to ground referring expressions and detect ambiguities, learning strong multi-modal representations in situated dialogues. We then analyse the signals required for models to learn to handle miscommunications, and propose a cross-modal taxonomy of clarifications to assess the contribution of distinct modalities. Our experiments with different model architectures and training objectives reveal that secondary objectives are essential to integrate multiple modalities (dialogue, visual and relational), leading to models better suited to deal with challenging clarifications in conversations. Finally, we evaluate how generative multi-modal LLMs handle both miscommunications and repairs by releasing a new benchmark, BlockWorld-Repairs, based on human data collections and studies. We then propose alternative training approaches that encourage models to learn from interactive settings, generalising to handling both instructions and subsequent repairs for successful task-completion. Throughout this thesis, we highlight the challenges posed by miscommunications and present approaches to develop robust collaborative conversational AI models better adapted for human interactions.Item Coupled stability of offshore wind monopile foundations and submarine slopes(Heriot-Watt University, 2025-04) Song, Benjian; Cummins, Doctor Cathal; Zou, Professor QingpingIn response to the current research gap concerning the existing of monopiles in submarine slope areas, this thesis conducts a study through finite element numerical simulation. The key methodologies employed include using a modified Mohr-Coulomb model with strain accumulation that considers the effects of high cycle accumulation strain for describing the mechanical behaviour of drained seabed sand, utilising the shear strength reduction method to determine the safety factor of submarine slopes, and employing a soil-pile interaction model to describe the contact behaviour between the seabed and the monopile. The study uses the Abaqus package to solve for multi-field coupling in porous media. The research primarily considers three main topics: the coupling effects of monopiles and submarine slopes under long-period cyclic loads and, under extreme storm load conditions, and a novel failure surface prediction method for submarine slopes according to seabed topography and finite element method. The study finds that the installed monopiles can act to prevent slope sliding, resulting in a slight increase in the overall safety factor of the slope. However, under high cycle accumulation strain and the influence of storms, an area around the monopile characterised by a funnel shape may experience abnormally high pore pressures and strain softening, leading to a reduction in the overall safety factor of the slope. Although the displacement of the monopile caused by storms or accumulation strain remains within a small range, the occurrence of a landslide could suddenly increase the displacement of the monopile, causing instability or even collapse of the monopile and tower structures. Considering the complex submarine topographical features potentially encountered in practice, a new finite element-based model for predicting potential failure surfaces of submarine slopes is proposed. This model demonstrates that the installed monopiles but before loading does not significantly affect the location of potential failure surfaces of submarine slopes, but it does influence the depth and safety factor of these failure surfaces.Item Facilitating assessment of topic modelling credibility(Heriot-Watt University, 2025-04) Liang, Yibo; Chantler, Professor MikeThis thesis investigates the factors influencing the perceived credibility of the Latent Dirichlet Allocation (LDA) algorithm, particularly focusing on the roles of transparency and interface design. Despite LDA’s widespread use in topic modeling within natural language processing, its opaque nature often leads to user distrust, impacting its application in areas requiring reliable decision-making support. The study employs a qualitative research approach, integrating user-based experiments and semi-structured interviews to explore how different levels of transparency and interface designs affect user perceptions of credibility. Experiments were designed to vary in the amount of algorithmic workings exposed to participants, alongside interface variations to assess their impact on user trust and understanding. Results indicate that while increased transparency generally enhances trust, overly complex information presentation can diminish the perceived credibility of the LDA algorithm. Moreover, intuitive and well-designed interfaces were found to significantly bolster user trust and comprehension, suggesting a delicate balance between transparency and usability is crucial for enhancing the credibility of algorithmic systems. This research contributes to the fields of human-computer interaction and machine learning by providing empirical insights into designing more effective user interfaces for complex algorithms. It highlights the importance of user-centered design principles in improving the accessibility and trustworthiness of machine learning models, with implications for developers and researchers seeking to foster greater acceptance and use of AI technologies.