AFArianna Fedeli

Arianna Fedeli

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Arianna Fedeli has a total of 37 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Automated Analysis of Crowdsourced Reports with Multimodal LLM Workflows: The Case of Citizen-Driven Waste Reporting

    by Maria Teresa Rossi, Arianna Fedeli, Ludovico Iovino, Amleto Di Salle, Leonardo Mariani

    Crowdsourcing platforms often rely on Human-in-the-Loop (HITL) workflows to validate submissions, assess report quality, detect duplicates, and assign gamification rewards. As the volume of multimodal reports grows, manual moderation can become a processing bottleneck, particularly in citizen-driven waste-reporting systems with limited operational resources.This paper investigates to what extent Large Language Model (LLM) workflows can support or selectively automate moderation activities in crowdsourced waste-reporting workflows. We study six multimodal LLM workflows configurations in the COBOL project, combining LLM-only zero-shot and few-shot settings and retrieval-augmented generation (RAG)-based zero-shot settings with LLaVA and Qwen2.5-VL. Using 309 real-world reports, we evaluate attribute assessment with correct, deliberately corrupted, and omitted structured waste metadata, retrieval-assisted duplicate detection, execution time, and selective-processing strategies.Performance varies substantially across attributes, configurations, and input conditions. Omitting structured waste-type and size values substantially reduces accuracy for most configurations. Deliberately corrupting these values produces heterogeneous effects, including accuracy losses for several configurations and waste-type improvements for the two LLaVA-based LLM-only configurations. These gains do not establish reliable correction of erroneous metadata. Duplicate detection exhibits low precision in the evaluated pipeline. LLM workflows execution times are substantially lower than the manual assessment times recorded during ground-truth construction. Retrospective agreement- and category-based analyses identify candidate subsets for selective attribute processing.The results motivate further evaluation of LLM-assisted moderation but do not establish reliable autonomous processing of complete reports. Privacy screening, duplicate handling, prospective validation of selective-processing policies, and end-to-end human-workload reduction require further investigation.

  • Open Access

    Assessing Capabilities of Large Language Models for Multi-Level Modeling

    by Arianna Fedeli, Ludovico Iovino, Maria Teresa Rossi, Adrian Rutle

  • Open Access

    Models@Orbit: Model-Driven Engineering for the New Space Era

    by Francesco Basciani, Arianna Fedeli, Ludovico Iovino, Patrizio Pelliccione

  • Open Access

    A model-driven framework for supervised reconfiguration of satellite schedulers

    by Francesco Basciani, Arianna Fedeli, Daniele Masti, Ludovico Iovino, Paolo Serri, Patrizio Pelliccione

    Satellite operations rely on schedulers that must satisfy strict timing, dependency, and resource constraints. In current industrial practice, these scheduler configurations are often defined manually and remain largely static after deployment, making updates labor-intensive and limiting adaptation to changing operational conditions. This paper presents a model-driven decision-support framework organized around a Digital Twin instance for the supervised runtime reconfiguration of satellite scheduler configurations in safety-critical systems. The framework combines an Ecore-based scheduler metamodel, graphical modeling support, EVL-based validation, telemetry-driven monitoring, simulation-assisted optimization, and interpreter-based generation of deployable scheduler configurations. In this way, the same scheduler model is preserved throughout design-time specification, validation, optimization, supervision, and runtime configuration generation, while keeping final deployment under human control to ensure certification constraints are met. We demonstrate the approach on an anonymized industrial scheduling scenario derived from a commercial satellite case provided by Thales Alenia Space Italia. We evaluate the framework through two complementary activities: a focus group with six TAS-I experts, used as an expert resonance study to assess usefulness, understandability, and industrial plausibility, and a toolchain-level validation of the MDE pipeline, measuring model loading, constraint checking, configuration generation, and traceability. The results indicate that the framework may reduce manual scheduling effort, support traceable and consistent reconfiguration decisions, and help engineers reason about controlled adaptation of scheduler configurations to runtime evidence without altering the certified switching logic of the onboard platform.

  • Open Access

    The Case Studies: Data Acquisition, Natural and Anthropic Risks Analysis, Network of Competences as a Tool for Preservation of the Culture Heritage

    by Roberto Aloisio, Ilia Antenucci, Maria Giovanna Brandano, Elena De Panfilis, Speranza Falciano, Arianna Fedeli, Ludovico Iovino, Marco Modica, Ugo Rossi, Enrico Stagnini

    The chapter embraces a variety of approaches to risk analysis and identification of the best ways to safeguard cultural heritage. None of the proposed methods is exhaustive but each contributes to formulating a multidisciplinary method which, as this volume repeatedly shows, the topic itself requires. The academic study that underlies each contribution does not forget applicative and operational implications, particularly in the area studied by the group: the city of L’Aquila. First of all, the case studies of the entire Spoke 7 are distinguished by types, identified according to different scales of reference. The AirTable platform, used by all Spoke Partners, becomes an opportunity for exchange and collaboration between those who, for example, operate on the same type of risk. The scale of territories and landscape is the object of analysis of aerial photogrammetry, an aid to the protection and documentation of cultural heritage, but also an important tool for urban planning and management. On a national scale, the impact of natural events on cultural heritage has been studied, resulting from the interaction between specific hazards and the intrinsic characteristics of individual artefacts, but also the communities’ capacity to absorb the impact of the event. Subsequently, we delve into the communities in a proactive approach to risk management, where we propose to collaborate with local communities in order to generate an endogenous capacity to face the risk, anticipating it, then working with awareness on prevention. The fil rouge is a constant eye on providing tools for managers of the res publica and decision makers, who can use georeferenced maps for planning, allocate resources for vulnerable communities, and use methodological frameworks to develop virtuous policies and forward-looking strategies for cultural heritage: the proposal that closes the chapter refers to the scale of the historic city, consisting of the planning and creation (in progress) of a network of expertise, cultural specialists and public administrations in the historic city of L’Aquila, to make possible shared and transversal policies—from decision makers to communities—for the care of cultural heritage, which is the identity and history of the communities that live it.

  • Open Access

    Noise-reduction-oriented super-resolution reconstruction for precision agriculture applications

    by Minh-Trieu Tran, Arianna Fedeli, Juan Antonio Piñera García, Patrizio Pelliccione, Thanh-Phuong Nguyen


  • A Research Roadmap for Digital Twins of an Organization

    by Arianna Fedeli, Ghina Kassem, Khaled Sherif, Emanuele Laurenzi, Andrea Polini

  • How low-code platforms support digital twins of processes

    by Arianna Fedeli, Amleto Di Salle, Daniela Micucci, Luciana Rebelo, Maria Teresa Rossi, Leonardo Mariani, Ludovico Iovino