157082 research outputs found
Sort by
R-RNet: Probability-Driven Networks for Pedestrian Trajectory Prediction
No full textAccurate prediction of pedestrian trajectories is crucial for safe motion planning of autonomous vehicles in urban environments. Many existing temporal and generative models are constrained by the long-tailed distribution of the data, which limits their ability to handle random or irregular pedestrian movements. Moreover, few studies have addressed the problem of scoring and probabilistic evaluation of predicted trajectories, despite their importance for downstream decision-making tasks. To address these issues, we propose a regularization-randomization network (R-RNet). The core regularization-randomization (R-R) module enables flexible trajectory prediction across diverse scenarios, while the probability predictor provides trajectory scoring and probability estimation to enhance reliability and utility in subsequent tasks. Besides, a self-attention mechanism is utilized to enhance prediction performance by capturing features from the distribution of the goals. The experimental results on the ETH and the UCY datasets show that R-RNet is capable of making reliable evaluations on output trajectories and achieves competitive results in terms of average displacement error and miss rate, while maintaining a lightweight architecture. Extensive experiments and analyses underscore the critical importance of both regularization and randomization operations
Optical Spin Effects Induced by Phase Conjugation at a Space‐Time Interface
No full textElectromagnetic temporal boundaries, emerging when the constitutive parameters of a medium undergo abrupt temporal variations, have garnered significant interest for their role in facilitating unconventional wave phenomena and enabling sophisticated field manipulations. A key manifestation is temporal reflection in an unbounded spatial domain, where a sudden temporal discontinuity induces phase‐conjugated backward waves alongside anomalous spin conversion. This study explores distinctive spin‐conversion dynamics at a time‐dependent spatial interface governed by Lorentz‐type dispersion, in which the plasma frequency undergoes rapid modulation over time. The interaction of a circularly polarized wave with a space‐time interface excites electromagnetic signals at the system's natural resonance, allowing precise control over polarization states. The scattered field stems from the combined influence of temporal and spatial boundaries, yielding a superposition of the original incident wave's polarization and its phase‐conjugated counterpart
Real-time data-driven multi-objective optimization in B2B marketing using digital twins
No full textIn the competitive B2B marketing environment, optimal decision-making requires intelligent and data-driven models that can respond to real-time market changes. This research presents an innovative framework for multi-objective optimization in B2B marketing that uses digital twins and real-time data to maximize profits, reduce marketing costs, and increase customer engagement. To achieve these goals, two algorithms, Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Whale Optimization Algorithm (WOA), provide a combination of high convergence speed and accuracy in finding optimal solutions. The results show that NSGA-II performs better when quick decision-making is required, while WOA provides more optimal solutions in some cases. Also, examining the role of digital twins showed that the proposed model can reduce additional costs and improve decision accuracy by continuously adjusting marketing strategies. Sensitivity analysis also confirmed that increasing marketing budgets and improving customer engagement rates directly impact growing profitability. The results of this research show that the proposed optimization framework, by integrating digital twins and multi-objective meta-heuristic algorithms, has the ability to improve resource allocation and performance indicators in dynamic and uncertain environments. This approach can be used in areas such as B2B digital marketing, supply chain management, and resource allocation in online service systems.</p
Categorising residential energy demand datasets in the UK
No full textAccess to high-quality residential energy demand data is crucial for research and policymaking. In the transition to a modern, digitalised energy system, datasets should be visible and accessible to end users. However, the absence of standardised data release guidelines and metadata standards creates challenges in data visibility, accessibility, and comparability. These challenges lead to repetitive and time-consuming searches for relevant datasets. This study examines twenty-four UK residential energy demand datasets, highlighting inconsistencies in how they are catalogued, documented, and structured. A novel classification scheme is introduced to systematically document dataset attributes, scope, and contextual variables. Subsequent categorisation of the twenty-four datasets using the classification scheme enhances data discovery, comparability, and consistency, while also identifying gaps. This article also addresses the evolving landscape of residential energy demand datasets and policy and the role of data in supporting efforts to decarbonise the building stock. This work highlights the need for greater standardisation and accessibility, emphasising the importance of harmonised metadata, improved documentation, and cross-dataset compatibility to support future research and policymaking
AutoMedTS: Automated Modeling of Physiological Time Series for Surgical Suturing Action Recognition
No full textIn laparoscopic surgical training and evaluation, real-time recognition of surgical actions with transparency outputs is crucial for automated, objective, and immediate instructional feedback to support skills improvement. However, we face challenges due to limited dataset sizes and variability in surgical environments. This study presents AutoMedTS, an end-to-end automated machine learning framework customized for medical time-series data, enabling rapid deployment using surgical suturing trajectories collected from both expert and novice surgeons. The proposed method features key improvements including: (i) a novel temperature-scaled Softmax resampling technique effectively addressing severe class imbalance, and (ii) an uncertainty-aware ensemble selection mechanism ensuring robust predictions across surgeons with varying skill levels. Additionally, the approach emphasizes model transparency to meet the high standards of reliability and transparency required in medical applications. Compared to deep learning methods, traditional machine learning models not only facilitate efficient rapid deployment but also offer significant transparency advantages. Experimental results demonstrate that our method provides fast, stable, and reliable real-time surgical action recognition in clinical training environments. Code and data are publicly available at https://github.com/baobingzhang/AutoMedTS
Perceptions and Experiences of the Menstrual Cycle amongst Elite Adult and Adolescent Football Players
No full textThe purpose of this study was to investigate players’ experiences and perceptions of the menstrual cycle (MC) and the perceived impact on performance. Female elite adult (n = 31, age 24.6 ± 5.1 years) and adolescent (n = 65, age 15.0 ± 1.1 years) players completed an online questionnaire consisting of quantitative and qualitative questions. MC symptoms were experienced by 90.1% naturally menstruating participants (86.9% adolescents and 93.6% adults (x2 = 1.53, df = 2, p = 0.47, n = 92)), and 78.3% adolescents perceived their MC impacts performance, compared to 96.4% adults (x2 = 4.54, df= 1, p = 0.033, n = 74). Physical symptoms, psychological symptoms, and energy levels were cited as key reasons for the MC negatively impacting performance. Challenges in communicating MC experiences were reported by 44.92% (n = 23) adolescents compared to 20.0% (n = 6) adults (x2 = 7.29, df = 2, p = 0.026, n = 82), with a perceived lack of knowledge, ability to relate and awkwardness cited as key reasons. Football players report wellbeing and performance impacts due to their MC, highlighting the need for individual understanding and support. Furthermore, understanding the experiences of adolescents enables the development of targeted support structures that equip them with tools to manage and communicate about their MC, and hopefully preventing issues as they become senior players
Artificial intelligence and machine learning for the diagnosis of Huntington disease: a narrative review
No full textBackground and Objective: Huntington's disease (HD) is a neurodegenerative disorder currently diagnosed by genetic tests and motor symptoms observation. However, these methods are either invasive or lack precision in diagnosing different stages, including presymptomatic states. These limitations have driven interest in the application of machine learning (ML) techniques to analyze patient data, identify HD patients, and uncover valuable biomarkers for diagnosis. Despite the growing body of research, a review of ML applications for HD diagnostics has been lacking. The review aims to provide a summary of ML methods used to diagnose HD and key diagnostics biomarkers that distinguish it from other neurodegenerative diseased (NDDs). Methods: A narrative review of English, peer-reviewed articles and conference papers that conducted experimental designs and employed ML or artificial intelligence (AI) algorithms for diagnostics. This includes those studies published from 2010 until 2023 on PubMed, IEEE and Heriot-Watt Discovery digital libraries. Amongst them, a total of 54 papers were found relevant and included in this review.Key Content and Findings: The review revealed the power of ML models for diagnosing HD from healthy controls, commonly by using physiological signals. Besides, decision tree-based models were the most used ML approach, offering a favourable balance between diagnostics performance and interpretability. Furthermore, despite that HD clinical scores emerged as crucual diagnostic features for identifying HD and discriminating them from control and other NDD conditions, more impactful features, such as brain structures, like caudate volume were found to improve the diagnosis. Conclusions: This review offers valuable insight for researchers and healthcare professionals, highlighting common ML applications for diagnosing HD, including data sources, modalities, preprocessing methods, and key biomarkers. Future research can refine diagnostic techniques by advancing from classical ML models to advanced approaches, leveraging state-of-the-art techniques, such as transformers to enhance performance, utilizing them for clinical decision-making, tailoring therapy development
The Prevalence and Potential Problem of Cuteness in Zoomorphic Robots
No full textCuteness is a powerful aesthetic, and psychological research shows that cute things such as infants, baby animals, and toys capture and secure our attention, promote nurturing behaviour, and influence our preferences. Therefore, cuteness is a common design outcome in many consumer products, including robotics. However, we suggest that making cute zoomorphic robots may not be without its issues due to the complexities introduced by the analogies they make to various animals. We summarise the impact of cuteness in animals and robotics and analyse the intersection of the two domains by comparing the presence of baby schema features in different canine zoomorphic robots and dog breeds. Finally, we speculate on the benefits and drawbacks to cute zoomorphic robots, and provide suggestions for a new design approach that centres animals’ well-being. The aim of this work is to synthesise research on cuteness from different disciplines and prompt robot designers to be more conscious of cuteness and its potentially detrimental consequences in zoomorphic robots
Access Repository Dashboard
Do you manage Heriot Watt Pure? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!