10081 research outputs found
Sort by
Vehicular-based support to cutting-edge application scenarios in next-gen networks
Get PDFThe rapid advancements in in-vehicle computing, communication, and software enable drivers to access diverse distributed applications and services. Edge Computing and frameworks like the European Telecommunications Standards Institute (ETSI) Multi-access Edge Computing (MEC) are intended to be key in standardizing service execution at the network edge. Complementing this, Vehicular Cloud Computing (VCC) and Vehicular Edge Computing (VEC) augment edge computational capacity by harnessing vehicle computing, storage, and communication resources. Together, these technologies create a resource continuum, enabling real-time processing, low latency, and improved service quality. This dissertation presents several solutions to contribute to the creation of an innovative and powerful vehicular-based support for edge applications in next-generation networks. Hence, it extensively studies and contributes to enhancing vehicular communication reliability and advancing the vehicular computing paradigm by developing simulated/emulated platforms for application design in this domain. The key contributions of this PhD dissertation include: i) a reputation-based system to exclude misbehaving vehicles within Vehicle-to-Vehicle (V2V) communications, ii) a dataset comprising V2V messages compliant with European standards, iii) an extensive study on the usage of a Time-Sensitive Networking (TSN) middleware to improve Quality of Service (QoS) and data delivery in Vehicle-to-Everything (V2X) communications, iv) the design of a vehicular computing architecture based on recognized standards (e.g., MEC) to promote interoperability and scalability, and v) a novel simulation tool allowing researchers and engineers to design, test, and enhance distributed applications that exploit vehicular, edge, and cloud computing concepts. Furthermore, this dissertation explores MEC deployment within 5G networks using Open Radio Access Network (O-RAN) specifications, establishing a foundation for a real-world vehicular cloud computing testbed. This led to an openly accessible middleware for developing near real-time applications following O-RAN standards, essential for integrating MEC and O-RAN. All proposed solutions have been rigorously tested, demonstrating their effectiveness in advancing next-generation vehicular and edge computing
Application of advanced car design methodologies and innovative simulation systems for the development of new low-environmental-impact vehicles
Get PDFThe automotive industry is at a crossroads. For nearly a century, physical prototypes, particularly clay and foam models, have been the cornerstone of the design process. While these models offer invaluable insights into aesthetics and functionality, their creation is an incredibly resource-intensive and environmentally taxing process. Due to its advantages, this method is difficult to replace, despite its downsides, and the challenge now is to find a way to move forward without losing the benefits that physical prototypes can provide. This doctoral research explores how the automotive industry can reduce its reliance on these traditional, resource-heavy methods by embracing digital tools and technologies. The goal is not to replace the tactile experience of a physical prototype, but to complement it with advanced digital solutions that can make the design process faster, cleaner, and more sustainable. However, these tools can only perform at their full potential when integrated into a robust methodological framework that enables them to operate in harmony and amplifies their strengths. By utilizing the IDeS framework as the foundational structure and incorporating associated methods such as QFD and SDE, this research proposes innovative methodological approaches to achieve these objectives with a circular approach to the design process. These tools are combined with modeling, simulation, and optimization software to create a digital ecosystem that enhances whole the design process while minimizing environmental impact. By harnessing the power of these technologies, the study demonstrates that it's possible to create highly detailed, immersive digital prototypes that replicate the aesthetic and functional qualities of their physical counterparts, but without the associated waste, cost, and environmental harm. In doing so, this research lays the groundwork for a new approach to automotive design, one where sustainability isn’t an afterthought but is woven into the fabric of the creative process from the very beginning
Race smart, last longer: deep learning approaches for li-ion battery state estimation and autonomous racing vehicles
Get PDFAs the demand for advanced autonomous systems grows, reliable energy storage and management becomes paramount, especially in high-performance contexts such as autonomous racing vehicles—including cars and drones. This dissertation explores two interrelated topics: (1) the accurate estimation and prediction of Lithium-ion (Li-ion) battery states, and (2) the development of Autonomous Vehicles (AVs) in competitive racing environments. The integration of these topics underscores the pivotal role that energy management plays in maximizing autonomous systems’ efficiency, safety, and performance. The first subject addresses the challenges associated with the state estimation of Li-ion batteries, which are the cornerstone of energy storage in modern autonomous systems. Accurate state estimation is critical for ensuring the longevity, reliability, and optimal performance of these batteries, particularly in applications where they are exposed to extreme operational stress. Through the application of deep learning, this research improves the accuracy of battery state estimation/prediction. The second subject focuses on the development of AVs in the context of racing. Racing provides a unique testing ground for autonomous systems, where rapid decision-making and precise control are critical. This research contributes to the field by developing reinforcement learning techniques for autonomous driving. Additionally, a novel dataset for autonomous drone racing is introduced, which provides a benchmark for high-speed navigation tasks. The third subject explores the potential for integration between these two domains, which is the unifying theme presented throughout the study. This part highlights the symbiotic relationship between battery performance and vehicle control policies' success. Overall, this dissertation makes contributions to both fields. It presents open-source software and publicly available datasets that support the research community in advancing these domains. These findings lay the groundwork for future research into the convergence of energy management and autonomous system design, promising further innovations in the pursuit of more sustainable and capable autonomous technologies
Machine learning methodologies for supporting HPC systems operations
Get PDFThe growing size and complexity of modern high-performance computing systems demand advanced data collection, monitoring, and machine learning methodologies for effective management and operations, collectively referred to in the literature as operational data analytics (ODA). The thesis introduces a comprehensive ODA framework addressing key challenges: open-ended data exploration, unsupervised anomaly detection, and long-term anomaly prediction. The first part of the comprehensive ODA framework is the methodology used to perform open-ended data exploration and analysis, called the DEM (data exploration model). DEM forms the foundation of the ODA framework, requiring no structured or labeled data, making it ideal as the first machine-learning model for HPC systems. It provides operational insights, helping administrators and stakeholders identify metrics for further analysis with specialized machine-learning models. The second component of the ODA framework is RUAD (Recurrent Unsupervised Anomaly Detection), a novel model that addresses the limitations of current state-of-the-art anomaly detection methods. Unlike traditional approaches that require labeled data or exhibit poor performance in unsupervised settings, RUAD outperformed all previous state-of-the-art semi-supervised and unsupervised techniques. RUAD achieves an AUC of 0.763 for semi-supervised and 0.767 for unsupervised training, surpassing the state-of-the-art method (AUC 0.747 semi-supervised, 0.734 unsupervised). It also significantly outperforms clustering-based unsupervised anomaly detection (AUC 0.548). The third component of the ODA framework, GRAAFE GRaph anomaly anticipation framework) extends anomaly detection to anomaly prediction using graph neural networks (GNNs). The physical layout of compute nodes in a compute room is modeled as a graph, with nodes as vertices and edges representing the physical distances between them. By leveraging spatial information ignored by per-node models, GRAAFE's GNN surpasses state-of-the-art anomaly prediction methods, achieving AUCs ranging from 0.91 to 0.78, compared to 0.64 to 0.5 from existing approaches. GRAAFE also pioneers long-term (over eight hours ahead) node failure predictions for high-performance computing systems
Moving towards Problem-Oriented Policing: what socio-criminological challenges for the Local Police of Ravenna?
Get PDFIl Problem-Oriented Policing (POP) è un approccio innovativo di polizia che si concentra sulla risoluzione dei problemi e sulla prevenzione delle cause alla base delle richieste di intervento. Integrato con il modello SARA (Scanning, Analysis, Response, Assessment), il POP mira a ridurre i reati, rafforzare il legame tra polizia e comunità e prevenire lo spostamento dei reati, favorendo un cambiamento culturale verso una gestione più inclusiva e orientata alla comunità.
Questo studio esplora l’applicabilità del POP in Italia, utilizzando la Polizia Locale di Ravenna come caso di studio. Il confronto tra le dinamiche italiane e i modelli adottati nei Paesi Bassi e nel Regno Unito evidenzia le lacune nella letteratura italiana riguardo alla cultura e ai modelli di polizia, mettendo in luce la distanza rispetto ad approcci innovativi presenti in altre nazioni europee. La metodologia ha seguito i passaggi del metodo SARA. Si è effettuata una mappatura e un’esplorazione flaneuristica delle zone problematiche della città, validata da interviste semistrutturate e analisi delle immagini fornite a cittadini, associazioni e poliziotti (scanning). I dati ottenuti sono stati analizzati con Nvivo per estrapolare gli assi tematici principali (analysis). Questi sono stati confrontati con quelli raccolti nei Paesi Bassi, attraverso interviste a rappresentanti locali e attori sociali (response). È stato proposto un design thinking a 30 operatori di Polizia Locale di Ravenna per illustrare il POP e co-costruire progetti di prevenzione. Infine, i dati sono stati analizzati e comparati con quelli ottenuti in Gran Bretagna, dove, con la Salford University, è stato partecipato al progetto POPLAR per rimodulare il processo valutativo delle attività di polizia (assessment). I risultati mostrano che il POP è efficace nella riduzione dei reati e nel miglioramento della relazione con la comunità, ma in Italia emergono resistenze culturali e organizzative che ostacolano l'adozione del modello.Problem-Oriented Policing (POP) is an innovative policing approach that focuses on problem-solving and preventing the underlying causes of intervention requests. Integrated with the SARA model (Scanning, Analysis, Response, Assessment), POP aims to reduce crime, strengthen the relationship between police and community, and prevent crime displacement, fostering a cultural shift towards a more inclusive and community-oriented management. This study explores the applicability of POP in Italy, using the Local Police of Ravenna as a case study. The comparison between Italian dynamics and the models adopted in the Netherlands and the United Kingdom highlights gaps in the Italian literature regarding police culture and models, revealing the distance from innovative approaches present in other European countries. The methodology followed the steps of the SARA model. A mapping and flâneuristic exploration of problematic areas of the city were conducted, validated through semi-structured interviews and analysis of images provided by citizens, associations, and police officers (scanning). The data collected were analyzed using Nvivo to extract the main thematic axes (analysis). These were then compared with data collected in the Netherlands, through interviews with local representatives and social actors (response). A design thinking workshop was conducted with 30 Local Police officers in Ravenna to illustrate the POP methodology and co-create prevention projects. Finally, the data were analyzed and compared with those obtained in the UK, where, in collaboration with Salford University, the process for evaluating police activities was restructured through participation in the POPLAR project (assessment). The results show that POP is effective in reducing crime and improving the relationship with the community, but in Italy, cultural and organizational resistances hinder the adoption of the model
Response of Vitis vinifera cultivars to multiple summer stresses and development of precision strategies to mitigate negative effects of climate changes on grape and wine
No full textThe thesis explores innovative strategies to mitigate the impacts of heatwaves on grapevine cultivation. It addresses the challenges posed by climate change and aims to develop tools to safeguard plant vitality and production. Three experiments are conducted to achieve this objective.
In the first experiment, conducted in pots, a multifunctional irrigation system is implemented to regulate the microclimate around the fruit-zone during heatwaves. Two red grape varieties, 'Sangiovese' and 'Montepulciano', are cultivated and exposed to different water regimes. Results show that misting increases yield and anthocyanin concentration, crucial for berry color.
A field experiment investigates misting's effect on a white grape variety over two years. Misting reduces air and berry temperatures, leading to increased yield and reduced sunburn damages. Flavonol synthesis remains unaffected by thermal reduction but is stimulated under increased light exposure.
In the third experiment, the effectiveness of a treatment with mineral compounds, such as chabasite-based zeolite, in enhancing anthocyanin accumulation and physiological performance is assessed. Results demonstrate the effectiveness of foliar zeolite treatment in reducing canopy temperatures and improving anthocyanin accumulation in grapes and wine.
In conclusion, the research provides insights into managing summer stresses in grapevines. Implementing these findings can aid in adapting grapevines to climate change challenges
Diagnostic performance of several indirect and direct biomarkers of fibrosis and fibroscan-based scores for the detection of liver disease severity in patients with metabolic dysfunction-associated steatotic liver disease (MASLD)
Get PDFBackground: La steatosi epatica associata a malattia metabolica (MASLD) comprende manifestazioni che variano dalla steatosi semplice (MASL) alla steatoepatite metabolica (MASH), che può progredire verso cirrosi e carcinoma epatocellulare. La MASLD è altamente prevalente globalmente, con le malattie cardiovascolari come principale causa di mortalità. Lo stadio della fibrosi epatica è un forte predittore di morbidità e mortalità, rendendo la sua valutazione cruciale per la stratificazione del rischio. Distinguere i pazienti con fibrosi avanzata rimane una sfida, poiché la biopsia epatica, pur essendo il gold standard, non è pratica per lo screening nella popolazione generale. I test non invasivi, inclusi metodi biochimici, antropometrici ed elastografici, mirano a predire la fibrosi avanzata.
Metodi: Uno studio prospettico monocentrico ha valutato la prevalenza di MASH grave e l’efficacia dei test non invasivi per la fibrosi avanzata nei pazienti ad alto rischio. Tra ottobre 2021 e maggio 2024, 84 pazienti, con sospettato di MASH severa, sono stati arruolati. I dati raccolti includevano misurazioni anamnestiche, antropometriche, biochimiche ed elastografiche tramite ultrasuoni, con tutti i pazienti sottoposti a biopsia epatica.
Risultati: MASH è stata riscontrata nel 72,6% dei pazienti, con il 52,4% con fibrosi almeno F2 e il 40,5% almeno F3. MASH grave (F3 e NAS ≥ 4) era presente nel 23,8%. La sindrome metabolica era presente nel 70,2%. Biomarcatori indiretti come GOT e APRI e punteggi complessi come FIB-4 hanno mostrato capacità diagnostica, ma il biomarcatore diretto PIIIP ha avuto una performance migliore (AUROC 0,75). La rigidità epatica e i punteggi AGILE hanno mostrato una modesta accuratezza diagnostica.
Conclusioni: Gli algoritmi diagnostici attuali, inclusi FIB-4 e rigidità epatica, sono meno efficaci nel rilevare MASH severa. Un algoritmo combinato di rigidità epatica e PIIIP ha mostrato maggiore accuratezza (81%), riducendo la necessità di biopsia nel 21% dei pazienti.Background: Metabolic-associated steatotic liver disease (MASLD) encompasses a range of liver manifestations, from simple steatosis (MASL) to severe forms like metabolic-associated steatohepatitis (MASH), which may progress to cirrhosis and liver cancer. MASLD is highly prevalent globally, with cardiovascular diseases being the primary cause of mortality. Liver fibrosis stage is a strong predictor of morbidity and mortality, making its assessment crucial for risk stratification and therapeutic interventions. However, distinguishing patients with advanced fibrosis remains challenging. Liver biopsy is the gold standard but is invasive and impractical for population screening. Non-invasive tests, including biochemical, anthropometric, and elastographic methods, aim to predict advanced fibrosis.
Methods: A single-center prospective study aimed to evaluate the prevalence of severe MASH and the effectiveness of non-invasive tests for advanced fibrosis in high-risk patients. Between October 2021 and May 2024, 84 patients were enrolled, all referred due to suspected severe MASH. Data collected included anamnestic, anthropometric, biochemical, and elastographic ultrasound measurements, with all patients undergoing liver biopsy.
Results: MASH was found in 72.6% of patients, with 52.4% having at least F2 fibrosis and 40.5% having at least F3 fibrosis. Severe MASH (F3 and NAS ≥ 4) was present in 23.8% of patients. The metabolic syndrome was highly prevalent (70.2%). Indirect biomarkers like GOT and APRI, and complex scores like FIB-4, showed some diagnostic ability, but direct biomarkers such as PIIIP had better performance (AUROC 0.75). Liver stiffness and AGILE scores demonstrated modest diagnostic accuracy.
Conclusions: The study shows that current diagnostic algorithms, including FIB-4 and liver stiffness, are less effective in detecting severe MASH in a high-risk population. A diagnostic algorithm combining liver stiffness and PIIIP demonstrated improved accuracy (81%), reducing the need for biopsy in 21% of patients
Prognostic and diagnostic markers in neuroendocrine thymic tumours:a pilot study
No full textBackground: Thymic neuroendocrine neoplasm (TNENs) are rare and aggressive cancers. The aim of the present study was to examine the prognostic significance of a panel of twenty biomolecular markers and to select the most common ones in order to better understand the pathogenesis and the prognostic factors of TNENs.
Material and methods: Clinical data and pathological tissue samples collected from surgically treated patients affected by TNENs were analysed. Twenty biomolecular markers, adopted in previously worldwide published papers, were utilized in a tissue microarray parallel in situ analyses. All patients were divided into three groups according to histological diagnosis: typical carcinoid(TC), atypical carcinoid(AC), and mixed (MxTNET). Overall and disease-free survival analysis was performed with the Kaplan-Meier method and the log-rank test.
Results: A total of 20 patients (1994-2012) was outlined from the databases.The great majority were male (90%) with a mean age of 55 years (IQR 35-82). 5- and 10- year OS of the cohort were 65% and 47%. 5-and 10-year DFS were 27% and 18%. Analysing the 20 markers, different expression between different histological types of TNETs was noticed with a significant trend in the DFS, towards two molecules such as Osteopontin (p=0.032) and PTEN (p=0.023). Analysing the overall survival, significant differences were noticed for pMTor (p=0.004). This protein had also an impact on DFS (p=0.002). Conclusion: According to the reported data, our analysis confirms the importance of OPN and PTEN also in neuroendocrine thymic tumours. Those parameters may be studied in larger cohort and integrated with the already known prognosticators to better define their impact on long term outcomes. Large multicentre studies are mandatory to define a scoring system able to predict the recurrence after radical surgery and define who benefit the most from aggressive therapie
Lexical complexity analysis from a computational perspective: a case study on institutional italian language
No full textThis study explores lexical complexity in Italian from a computational perspective, aiming to develop tools to analyze and manage it automatically. To this end, linguistic parameters influencing speakers' perception of complexity were selected, including morphological features. This led to the development of a morphological segmenter capable of decomposing words into roots and affixes, allowing extraction of quantitative morphological features. These features were incorporated into a model for identifying complex words in Italian. Given the lack of specialized datasets for this purpose, a dictionary of simple and complex words was created, and various linguistic feature configurations were tested to train the optimal statistical classifier. The classifier was then evaluated on a corpus annotated by humans, showing an accuracy comparable to human evaluations and validating the approach.
The system was applied to Cov-I-Cor, a corpus of institutional texts on health emergency management, constructed for this research. From Cov-I-Cor, a list of “pseudo-technical terms” that could be simplified was extracted. A comparative analysis with CORIS, a corpus of general Italian, highlighted lexical differences: CORIS showed greater lexical diversity, but Cov-I-Cor exhibited significantly higher informational density, suggesting institutional texts contain a higher proportion of content words. This finding is supported by the grammatical distribution, showing a higher noun-to-verb ratio in Cov-I-Cor, indicative of a more descriptive and informative nature. Additionally, institutional texts refer more frequently to concrete elements, aligning with their purpose of precise communication in regulatory contexts.Questo studio esplora la complessità lessicale da una prospettiva computazionale, con l’obiettivo di sviluppare strumenti automatici per la sua analisi e gestione. A tal fine, sono stati selezionati diversi parametri linguistici che influenzano la percezione della complessità da parte dei parlanti. Tra questi, alcuni riguardano la morfologia; pertanto, è stato addestrato un segmentatore morfologico automatico per scomporre le parole e ricavare parametri morfologici quantitativi. Questi ultimi sono stati inclusi nel primo modello di identificazione delle parole complesse per la lingua italiana.
In assenza di dataset specifici, è stato creato un dizionario di parole semplici e complesse, e sono state esplorate diverse configurazioni di caratteristiche linguistiche per addestrare il miglior modello statistico di identificazione delle parole complesse. Questo classificatore, valutato su un corpus annotato manualmente, ha dimostrato una precisione comparabile a quella delle valutazioni umane.
Il sistema è stato poi applicato al corpus Cov-I-Cor, una raccolta di testi istituzionali relativi alla gestione dell’emergenza sanitaria, per individuare una serie di "pseudotecnicismi" utili alla semplificazione del linguaggio. È stata inoltre condotta un’analisi comparativa tra Cov-I-Cor e CORIS, un corpus di italiano scritto generico, per evidenziarne le differenze lessicali. I risultati mostrano che, sebbene CORIS presenti una maggiore diversità lessicale, i parametri legati alla densità informativa sono significativamente più elevati in Cov-I-Cor. Questo indica che i testi istituzionali tendono a contenere una maggiore proporzione di parole contenuto rispetto al linguaggio generico, risultando più informativi. Tale caratteristica è confermata anche dalla distribuzione delle categorie grammaticali, con un rapporto più elevato tra sostantivi e verbi, che sottolinea la natura descrittiva e informativa dei testi istituzionali. Inoltre, questi ultimi si riferiscono più frequentemente a elementi concreti e reali rispetto ai testi generici
The role of legal technology in the harmonisation of EU law
Get PDFThis dissertation explores the transformative impact of legal technology, particularly natural language processing and knowledge modelling, on the analysis of legal harmonisation within the European Union. From a legal-linguistic perspective, this first involves analysing the key constituents of the phenomenon, including multilingualism, the nature of language and legislative sources, and legal interpretation, all of which play a pivotal role in shaping the harmonisation landscape. The core of the thesis, however, encompasses various hybrid legal technology methodologies for effectively analysing legal harmonisation within the European Union. The dissertation illustrates impactful applications through case studies, particularly in the context of EU-funded projects that enhance judicial cooperation across European Member States. By bridging legal and technological domains, the thesis contributes to understanding how digital tools, when implemented within robust methodological frameworks, can concretely enhance EU legal harmonisation, especially through the most recent generative AI technologies. Moving from traditional vector-based computational pipelines to network analysis, ontological modelling, and language models-based approaches, this research proposes methodologies that, while accounting for the complexity of the phenomenon, can have a meaningful, concrete impact on the workload and capacity of European public administrations, with additional positive effects on legal practice. The outcome of this research, conducted in the rigorous legal domain where results must meet the highest standards, is further validated by legal experts through manual knowledge modelling and their validation and discussion of the results. The conclusions assess the challenges and prospects of deploying Large Language Models (LLMs) and related technologies in analysing legal harmonisation, while also evaluating research findings. These cutting-edge technologies hold promise for accelerating the comparison and exploration of multilingual legislative texts. Despite these advancements, incorporating accurate, manually-crafted data remains essential. This approach is crucial for minimizing the risk of producing misleading and inaccurate information