Thesis Proposals

The starting point of this project is represented by the idea of designing and implementing a Web-GIS platform aimed at supporting infrastructure asset management. The case study will focus on a new metro line project, which provides a concrete context for the work. The platform is intended to serve as a decision-support tool that can be used throughout the entire life cycle of the assets, from design and construction to operation and maintenance. Its main purpose is to collect, process, and visualize heterogeneous data from multiple technological and non-technological sources. On this basis, the system will compute and monitor fifteen key performance indicators (KPIs) that describe infrastructure performance and sustainability. The work will focus on developing the complete workflow supporting the Web-GIS platform, from multi-source data extraction and ingestion to storage, KPI computation, and web-based visualization. A central database will act as the backbone for organizing heterogeneous inputs (manual data, IFC/BIM, questionnaires, GIS layers, IoT sensors, images, and video), while dedicated processing services will compute and update fourteen KPIs and make results available via APIs for GIS visualization and dashboarding. An essential aspect of the system will be its ability to handle temporal updates. This means that new incoming data should be incorporated into the database and reflected in the recalculation of KPI values, while at the same time maintaining a historical record that enables long-term monitoring and comparative analysis. This functionality is fundamental for capturing the evolution of infrastructure conditions over time. The expected outcome is a working Web-GIS platform prototype integrating data ingestion, database storage, KPI computation, and web-based GIS visualization. The implementation will be demonstrated on a sample set of KPIs, with an extensible architecture enabling the future inclusion of the complete KPI set.

This thesis focuses on the mathematical verification of "jets"—optimized C-language implementations of specific functions within the Simplicity language. The candidate will select a non-trivial cryptographic function, such as jet::bip_0340_verify (Schnorr signatures), and develop a machine-checked proof using a theorem prover like Coq or Lean. The objective is to demonstrate that the optimized jet is functionally equivalent to the verbose Simplicity program specification, ensuring both performance and high-assurance security. Expected Deliverables: ● A complete, machine-checked proof script in Coq or Lean ● Technical documentation outlining the proof strategy and equivalence results

Deep-learning systems can be accurate on average and still fail on individual medical images. The cause of failure matters: an examination may be degraded by motion or noise, the image may be intrinsically ambiguous even for expert readers, or the case may differ from the training distribution. Current uncertainty-quantification methods usually return a confidence or risk score that is useful for error detection, calibration, selective prediction, or out-of-distribution (OOD) detection. However, the same level of predictive uncertainty can arise from different failure sources, which may require different responses. Goal and objectives The thesis will investigate whether uncertainty can be made failure-source-aware and actionable. The objective is to test whether observable uncertainty signals contain enough information to distinguish clinically meaningful failure factors beyond generic predictive risk. - Build a benchmark that separates acquisition degradation, reader disagreement / intrinsic ambiguity, and distribution shift - Compare standard uncertainty signals: softmax confidence/entropy, calibration scores, deep ensembles, Monte-Carlo dropout, test-time augmentation, and representation/OOD scores. - Evaluate failure-source attribution at matched predictive risk: cases with similar error risk should still be distinguishable by the underlying factor. - Test whether factor-aware decisions improve simple intervention policies, such as reacquisition/quality control, expert review, or OOD referral, under the same intervention budget. Project environment, supervision and international mobility SafeMed is a collaborative research project between GIPSA-lab in Grenoble and the Bio-AI Lab at UiT – The Arctic University of Norway in Tromsø. The activities are based in GIPSA-lab, Grenoble, and are expected to start in February 2027.

Context: The increasing integration of advanced technologies in various industries necessitates robust validation methods for computer vision systems. A recent initiative aims to leverage generative AI to create a controllable simulator that can enhance the validation of face recognition systems. This approach addresses challenges such as environmental lighting variability and changes in user appearance, which are critical for ensuring the accuracy and reliability of biometric systems. Project description: This project will explore the use of generative AI as a controllable simulator for validating computer vision systems, with a particular focus on face recognition technologies. The research aims to evaluate the effectiveness of generative models by examining various aspects such as semantic fidelity, which includes identity and scene coherence; geometric fidelity, focusing on pose and multi-view consistency; and photometric fidelity, which analyzes lighting and lens effects. Additionally, the project will assess physical fidelity through the control of camera parameters, ground-truth reliability by measuring the accuracy of measurements, and dataset quality by emphasizing diversity and scalability. Objectives: • To evaluate the effectiveness of genAI in creating controllable synthetic data for computer vision validation. • To establish a comprehensive evaluation framework based on the proposed taxonomy. • To generate a dataset that meets requirements for effective validation of face recognition systems. • To explore potential applications of the findings in other industries, including automotive and beyond.

The thesis focuses on the design and development of a methodology for creating realistic animated 3D urban environments. The proposed system should enable users to generate topologically consistent urban scenes populated with various types of vehicles (e.g., cars, buses, bicycles, trucks, etc.), as well as pedestrians and animals, both static and dynamic, to achieve a realistic and comprehensive representation of an urban environment. The methodology should also define camera viewpoints consistent with those commonly found in major AI datasets for urban scene understanding. The resulting scene will be used to generate synthetic image datasets for training artificial intelligence models.

The proposed thesis focuses on the design and development/evolution of an immersive Virtual Reality (VR) simulation for training CBRN (Chemical, Biological, Radiological, and Nuclear) personnel in the medical field. The aim of the thesis, which will be carried out in collaboration with Fondazione LINKS, is to create a virtual environment in which personnel can practise the collaborative management of complex scenarios requiring the coordinated execution of medical procedures in the presence of contamination risks. The simulation will be developed as a multiplayer experience, allowing multiple participants to take part simultaneously, with particular emphasis on collaboration, coordination, and the correct execution of procedures. The activities will be developed and validated in collaboration with experts in the field of aeronautical CBRN operations to ensure that the virtual scenarios are consistent with real-world procedures and operating conditions. The experimental activities will also assess the possibility of integrating recently acquired, next-generation CBRN measurement instruments, devices, and equipment into the simulation, exploring their use and potential within emergency management and response procedures. The thesis may be carried out as a "thesis in a company" and, if deemed useful by the candidate, the possibility of also arranging an internship may be explored.

This thesis will concentrate on defining, analyzing, and validating figures of merit for advanced sensor fusion architectures within smart environments. The student will review performance and reliability metrics combining contactless spatial data (millimeter-wave radar sensors) with biometric and motion streams from low-power wearable devices (BLE/Zigbee). The work will propose a coherent framework of metrics tailored to real-time edge computing workflows, optimizing event detection while minimizing false alarm rates. Several composite objectives will be designed, for instance balancing system latency, data consistency, and user privacy constraints. Using simplified sensor simulators, data fusion models, or hardware testbeds, the thesis will explore trade-offs between different filtering policies and how edge computing architectures influence detection accuracy. The expected contribution is a clear metric and algorithmic framework to evaluate multi-sensor fusion strategies in home automation. Notes: The thesis will be carried out in a hybrid mode (Smart Working for software and code development) with periodic in-person testing sessions and laboratory activities at the Alessandria campus. Contact Patrik Resta resta.patrik83@gmail.com

This thesis aims to evaluate and compare two different approaches for the synthetic dataset generation in the car domain: CARLA2real (based on Unreal Engine) and NVIDIA Cosmo (a physical AI based-pipeline to support the photorealistic transfer).

This thesis explores the design and development of a prototypal framework for integrating multimodal perception, reasoning models, and agent orchestration in XR applications. These capabilities should be exploited to build spatially aware, multimodal AI agents that deliver low-latency, context-aware assistance in XR experiences. The system should make use of the most recent available AI models and frameworks, such as NVIDIA XR AI, exploring three core concepts: • Ingest real-world signals from XR devices, including video, audio, depth, pose and sensor data. • Connect agents to specialized tools and services for video search and summarization (VSS), for visual AI and video understanding, and for knowledge retrieval and retrieval-augmented generation. • Support multiple AI models.

Problem: The introduction of XR technologies is presenting unexplored challenges to final viewers. It is not clear where traditional cinematic techniques would also apply to immersive movies and their effects on the users' experience. Objective: Deepening existing studies in the context of cinematic AR/VR to cope with challenges regarding users' behavior in immersive movies, i.e., camera movements, watching conditions, scene cut techniques, fear of missing out, volumetric videos, lighting conditions, dynamic storytelling. Approach: Design studies and realize short immersive movies aimed to validate cinematic techniques and users' behavior when immersed in the watching experience Expected Impact: Enhanced experience for users watching immersive movies

Organizations increasingly operate in volatile, uncertain, and information-dense environments where effective decision-making depends on the ability to rapidly synthesize large volumes of data, align diverse perspectives, and coordinate action across distributed teams. Traditional decision-support systems, however, remain largely static and non-immersive, offering limited mechanisms for real-time collaboration, strategic alignment, and cognitive workload management. Despite significant advances in Artificial Intelligence, Virtual Reality, and Augmented Reality, the integration of XR technologies with adaptive management tools capable of actively supporting collaborative decision-making processes remains largely unexplored. This thesis aims to investigate how intelligent XR environments, integrating AI, VR, and AR, can enhance collaborative decision-making, strategic coordination, and sensemaking in complex organizational contexts. The research will explore how adaptive immersive systems can dynamically personalize information presentation, simulation scenarios, and communication strategies to improve team coordination, reduce cognitive overload, and facilitate consensus-building and negotiation among decision-makers. The student will design and develop a collaborative XR platform that combines immersive analytics, virtual agents, and AI-driven adaptive interfaces. The system will leverage multimodal behavioral signals, including gaze patterns, spatial movement, interaction dynamics, and response timing, to continuously assess collaboration quality, cognitive load, and individual engagement. Based on these inferences, the platform will dynamically personalize visualizations, recommendations, and decision-support strategies in real time, enabling teams to navigate complex information spaces more effectively. Particular attention will be devoted to the design of adaptive mechanisms for group sensemaking, the modeling of coordination signals, and the integration of human–agent and human–human decision loops within immersive environments. The developed solutions will be evaluated through user studies combining objective metrics, such as decision quality, coordination efficiency, task completion time, and interaction patterns, with subjective assessments of cognitive load, strategic alignment, perceived collaboration effectiveness, and user experience within the XR environment. The expected outcome of this research is the identification of design principles and adaptive strategies for XR-based decision-support systems that actively enhance collaborative intelligence in organizational settings. The findings are expected to advance the fields of Human-AI collaboration, immersive analytics, and strategic teamwork, informing the development of next-generation decision environments in which intelligent agents and adaptive interfaces play a central role in supporting complex, high-stakes organizational decisions. References: 1. B. Han et al., "Exploring Mediation by an Embodied Virtual Agent in Immersive Triadic Collaborative Decision-Making," in IEEE Transactions on Visualization and Computer Graphics, doi: 10.1109/TVCG.2026.3679103. 2. D. Garkov et al., "Collaborative Problem Solving in Mixed Reality: A Study on Visual Graph Analysis," in IEEE Transactions on Visualization and Computer Graphics, doi: 10.1109/TVCG.2026.3671472.

The increasing adoption of immersive technologies such as Augmented Reality (AR) and Virtual Reality (VR) is enabling new forms of remote and distributed collaboration. In many emerging scenarios, users operate across different realities—for instance, one participant interacting through AR while another is immersed in VR. These cross-reality settings introduce inherent asymmetries, as users may have access to different visual cues, perspectives, and interaction capabilities. Such differences can significantly influence how individuals coordinate, communicate, and contribute to shared tasks. This thesis aims to investigate the emergence of leadership in cross-reality collaborative environments. In particular, the work will explore whether leadership dynamics are primarily shaped by the technological medium (e.g., AR vs. VR) or by the level of mutual awareness between users, often referred to as co-presence awareness. Understanding these factors is essential for identifying how control, initiative, and decision-making are distributed in asymmetric collaborative settings. The student will design and implement experimental scenarios in which an AR user collaborates with a VR user on shared tasks. The experimental design will systematically vary the level of awareness cues available to participants, such as visibility of actions, gaze direction, or attention indicators. User interactions will be analyzed to assess leadership behaviors, including initiative taking, guidance, and task control. Both qualitative and quantitative methods will be employed to evaluate how different configurations impact collaboration dynamics. The expected outcome of this research is a deeper understanding of how immersive systems can unintentionally introduce imbalances in collaborative interactions. The findings will contribute to the design of cross-reality experiences that promote fair participation, balanced decision-making, and effective collaboration, regardless of the underlying technology. More broadly, the work may inform guidelines and design principles for developing equitable and user-centered XR collaborative systems. References: 1. The Effects of Sharing Awareness Cues in Collaborative Mixed Reality https://pmc.ncbi.nlm.nih.gov/articles/PMC7805624/  2. Empowerment and embodiment for collaborative MR systems https://doi.org/10.1002/cav.1838 3. Who Owns What? Psychological Ownership in Shared AR https://www.sciencedirect.com/science/article/abs/pii/S107158192100029X 4. Exploring User Behaviour in Asymmetric Collaborative Mixed Reality https://dl.acm.org/doi/abs/10.1145/3562939.3565630 5. Persuasive Vibrations: Effects of Speech-Based Vibrations on Persuasion, Leadership, and Co-Presence During Verbal Communication in VR https://ieeexplore.ieee.org/abstract/document/10108422 6. Towards Cross-Reality Interaction and Collaboration: A Comparative Study of Object Selection and Manipulation in Reality and Virtuality https://ieeexplore.ieee.org/abstract/document/10108776

xtended Reality (XR) technologies are opening new frontiers for the study of human cognition, emotion, and social behavior by enabling immersive, controllable, and interactive environments. Despite their strong potential, XR systems are still underutilized in psychological research, which traditionally relies on less dynamic and less ecologically valid experimental settings. As a result, fundamental questions about how immersive experiences shape mental processes, perception, and behavior remain only partially explored. This thesis aims to investigate how XR can be used as a tool to study and influence human experience, focusing on the role of adaptive and embodied virtual agents, immersive environments, and psychologically-driven interventions. In particular, the work will explore three complementary research directions: (i) the impact of virtual agent embodiment on user engagement, social presence, and self-disclosure; (ii) the use of eye-tracking data to analyze attentional patterns and predict perceived immersivity; and (iii) the effects of cyber-hypnosis and immersive suggestion techniques on arousal and subjective experience. The student will design and implement XR experimental scenarios integrating virtual agents, interactive environments, and adaptive mechanisms driven by multimodal data. Behavioral and physiological signals (such as gaze, interaction patterns, and biosignals) will be used to personalize the experience in real time. User studies will be conducted to evaluate the impact of different experimental conditions, combining objective measures with subjective assessments of presence, engagement, emotional response, and perceived immersion. The work will be carried out in collaboration with the University of Florence (UniFi), Department of Psychology, ensuring interdisciplinary integration between technological development and psychological research. The expected outcome of this research is a deeper understanding of how immersive technologies influence human experience and behavior. The findings are expected to contribute to both methodological advancements in XR-based psychological research and practical guidelines for designing immersive systems in domains such as training, therapy, and human-centered applications. More broadly, the thesis will help establish XR as a systematic and reliable tool for investigating psychological and social processes in controlled yet ecologically valid environments. References: Bosta, A., Vosinakis, S. (2026). Assessing the Immersive Experience with Physiological Measures: A Systematic Literature Review. In: Michael-Grigoriou, D., Zachmann, G., Kopper, R., Yoon, S.H., Zollmann, S., Bourdot, P. (eds) Virtual Reality and Mixed Reality. EuroXR 2025. Lecture Notes in Computer Science, vol 16101. Springer, Cham. https://doi.org/10.1007/978-3-032-03805-0_10

The integration of Large Language Models (LLMs) with Virtual Reality (VR) offers promising opportunities to redefine how users interact with immersive environments. This thesis explores the use of LLMs to support more natural, intuitive, and hands-free interactions in VR, moving beyond traditional input methods like physical controllers. The primary goal is to design and implement a framework that leverages the language understanding capabilities of LLMs to enable voice-driven interactions within VR environments. Potential use cases include hands-free locomotion (e.g., teleportation or virtual walking via spoken commands), interaction with virtual objects or interfaces through natural language, and the integration of multimodal LLMs to interpret visual inputs, such as referencing objects in the environment through verbal cues. The work will involve defining the system architecture, developing prototypes using Unity and voice processing tools, and exploring how LLMs can interpret and respond to user input in real time. The system will be evaluated through user studies, collecting both quantitative and qualitative data to assess usability, command recognition accuracy, response times, and overall user experience. References: [1] D. Calandra, F. G. Pratticò, and F. Lamberti, "Comparison of Hands-Free Speech-Based Navigation Techniques for Virtual Reality Training," in 2022 IEEE 21st Mediterranean Electrotechnical Conference (MELECON), Palermo, Italy, 2022, pp. 85-90. doi: 10.1109/MELECON53508.2022.9842994.  [2] J. A. V. Fernandez, J. J. Lee, S. A. S. Vacca, A. Magana, B. Benes, and V. Popescu, "Hands-Free VR," arXiv preprint arXiv:2402.15083, 2024  [3] S. Özdel, K. B. Buldu, E. Kasneci, E. Bozkir “Exploring Context-aware and LLM-driven Locomotion for Immersive Virtual Reality” arXiv preprint arXiv:2504.17331, 2025

Problem: Traditional Air Traffic Control (ATC) systems can lead to high cognitive workload, especially with increasing air traffic density and the introduction of new air vehicles like drones. Current visualization methods may not fully support complex trajectory planning and real-time monitoring. Objective: Augmented Reality (AR) tactical table for ATC that enhances situational awareness, supports efficient trajectory planning and monitoring, and includes Early Warning Systems (EWS). Approach: Develop an AR-enabled tactical table visualizing the urban airspace, including all aircraft (traditional and Urban Air Mobility), their trajectories, and potential conflicts. Integrate semi-automatic processes to assist controllers, reducing cognitive workload while maintaining human oversight. Multiple theses are available under this project: - AI/ML backend only: Design and develop the EWS system based on ML and Foundational Models (based on Radio Communications, ATC Visualization systems (radar, GPS), etc. - XR front-end/dashboard: Design and develop the XR visualization Expected Impact: Improved decision-making speed and accuracy; reduced cognitive workload for air traffic controllers; enhanced safety through advanced EWS; efficient management of complex and dense urban airspaces. References: https://ieeexplore.ieee.org/document/10764466, https://ieeexplore.ieee.org/document/9089606, https://dl.acm.org/doi/10.1145/3385378.3385380