Reserved topic scholarships
Intellectual Property Notice for PhD candidates under the UniTrento-FBK Agreement
Please read the following information carefully before submitting your application.
Intellectual Property of Research Results.The intellectual property rights of research results generated by PhD students under scholarships within the UniTrento-FBK Agreement shall belong to FBK.
Transfer of Intellectual Property Rights. FBK will establish agreements with PhD students regarding the transfer of intellectual property rights related to their research results.
Collaboration with UniTrento. If UniTrento academic staff contribute to research results obtained through PhD scholarships funded by FBK, the determination of IP shares will be defined through separate written agreements based on each party’s contribution. PhD students are required to collaborate with UniTrento in all necessary activities related to the joint management of IP.
Department of Information Engineering and Computer Science
A1 - Geometry-Consistent Open-World Foundation Models for Multi-Agent 3D World Modeling (Project: GUIDANCE (Debugging Computer Vision Models via Controlled Cross- modal Generation) FIS2023-03251; CUP E53C25000420001) (1 grant)
This project aims to enable scalable, robust, and platform-agnostic 3D world modeling by investigating whether geometry- consistent representations can be learned directly from multi-agent observations, without requiring explicit pose estimation as a
preprocessing step. Autonomous systems increasingly operate in heterogeneous, open-world environments where data are collected by multiple agents such as cars, drones, and scooters. Traditional 3D reconstruction pipelines rely on accurate
camera pose estimation (e.g., via Structure-from-Motion or SLAM), which becomes unreliable under extreme viewpoint changes, large baselines, dynamic scenes, or cross-platform sensing. These limitations constrain the robustness and scalability
of modern methods, including Gaussian Splatting approaches that assume precise pose supervision. By bridging multi-view geometry with foundation-scale learning, this project seeks to overcome these challenges. This research is supported by the
project GUIDANCE (Debugging Computer Vision Models via Controlled Cross-modal Generation) FIS2023-03251 – CUP E53C25000420001.
Contact: Niculae Sebe niculae.sebe@unitn.it
A2 - Reliable Open-World Video Models under Novelty and Ambiguity (Project: GUIDANCE (Debugging Computer Vision Models via Controlled Cross-modal Generation) FIS2023-03251; CUP E53C25000420001) (1 grant)
This project will develop task-aware open-set generation and scalable video world models that remain reliable under novelty, long time horizons, and ambiguous instructions. Building on prior foundations, concretely, the project will pursue three tightly
coupled thrusts: 1) upgrading open-world video perception from anomaly detection to LLM-guided concept proposal for unknown categories, 2) enabling open-set instruction understanding so an LLM can infer intent and respond with calibrated,
risk-aware reasoning under underspecified requests, and 3) building video-generation-grounded world models with long-horizon coherence to support object permanence, revisits, and controllable rollouts. This research is supported by the project
GUIDANCE (Debugging Computer Vision Models via Controlled Cross-modal Generation) FIS2023-03251 – CUP E53C25000420001.
Contact: Niculae Sebe niculae.sebe@unitn.it
A3 - Towards Explainable Foundation Models: Representation, Attribution, and Model Understanding (Project: GUIDANCE (Debugging Computer Vision Models via Controlled Cross-modal Generation) FIS2023-03251; CUP E53C25000420001) (1 grant)
The research activity will focus on explainability for foundation models, with particular emphasis on understanding, interpreting, and improving the transparency of large-scale pretrained models. The research will investigate methods to analyze how
foundation models represent knowledge, make decisions, and generalize across tasks and domains. The candidate will develop novel techniques for interpretable representations, attribution, and diagnostic evaluation of model behavior. The project will
combine theoretical advances with empirical validation on real-world computer vision datasets. This research is supported by the project GUIDANCE (Debugging Computer Vision Models via Controlled Cross-modal Generation) FIS2023-03251 – CUP
E53C25000420001.
Contact: Niculae Sebe niculae.sebe@unitn.it
B1 - AI-based XR systems for real-time interaction in the Musical Metaverse (Project: UE HE MUSMET – TURCHET; CUP E63C24002360006) (1 grant)
The candidate will work on the design and implementation of MR platforms supporting real-time multi-user musical interactions, with a particular focus on Mixed Reality scenarios where musicians will perform with real instruments and audiences engage simultaneously across physical and virtual spaces. A key area of exploration will be the use of generative AI for 3D content generation to enable dynamic environments and avatars generation within networked immersive platforms.
Contact: Luca Turchet luca.turchet@unitn.it
B2 - Design and evaluation of XR ecosystems for real-time interactions in the Musical Metaverse (Project: UE HE MUSMET – TURCHET; CUP E63C24002360006) (1 grant)
The candidate will work on the design and evaluation of radically new XR ecosystems that support real-time interactions in the Musical Metaverse. The interactions will occur between geographically distributed users in both virtual and mixed reality settings: between musicians, between musicians and audiences, and between audience members. Design and evaluation activities will be based on methodologies of the human-computer interaction field. Software development activities will leverage game engines and music information retrieval libraries. The activities will be conducted within the context of the European project MUSMET. Knowledge of music technology methods and tools is mandatory.
Contact: Luca Turchet luca.turchet@unitn.it
B3 - Study and development of “software-defined” technologies for the information transmission in 6G non-terrestrial networks (Project: MUR - PNRR - BaC RESTART INFINITE Sacchi; CUP D93C22000910001) (1 grant)
The new mobile radio communication standards that fall within the scope of the Sixth Generation (6G) provide for ever closer integration between terrestrial connectivity and space connectivity, the latter term referring to all those network segments that operate in the atmosphere (drones, stratospheric platforms) and beyond the atmosphere (low-orbit and geostationary satellites). The latest developments in 6G standardization point to a truly epoch-making shift: from the integration of terrestrial and non-terrestrial networks, seen as separate entities (5G), to the design and development of an integrated ecosystem, where the terrestrial and non-terrestrial components are considered in a native and interpenetrated manner (“Native NTN”).
Recently, the ITA-NTN (‘Integrated Terrestrial-Non Terrestrial Network’) research project, funded under the PNRR RESTART program (2023-2026), and its cascade project INFINITE (‘An Integrated and Sustainable Terrestrial/Non-Terrestrial Ecosystem for Anytime/Anywhere 6G Connectivity’) (2024-2025) have moved in this direction: studying a new technological approach to incorporate terrestrial and non-terrestrial network transmission and management technologies into an integrated ecosystem. The keyword in this process is ‘softwarization’, which means the implementation and management of all network levels, starting from the physical level, entirely in software manner.
The research project to be carried out during the three years of the PhD program builds on the concepts developed and results obtained in the ITA-NTN and INFINITE projects, of which it is a follow-up activity. The aim of the project is to create a software-defined environment for reconfigurable transmission in NTN networks, where the transmission format is decided adaptively based on service quality requirements and network conditions (propagation, mobility, congestion, multi-user). The initial conceptual diagram, extracted from [1] and shown in Figure 1 below, is based on the synergy between the modular and reconfigurable software-defined radio (SDR) implementation of the physical layer and an orchestrator agent based on software-defined networking (SDN) and artificial intelligence (AI) technologies for selecting the most suitable transmission format for the application context.
The research activity within the doctoral program will address two specific points:
• Implementation and emulation of the modular SDR library, containing the instantiations of the various software modules to create the different waveforms used in 6G non-terrestrial transmission (from basic OFDM to advanced multicarrier modulations, such as CE-OFDM, OTFS, OTSM, orthogonal and non-orthogonal multiple access strategies), implemented using open-source software tools (GNU-radio) and laboratory testing on low-cost hardware platforms for SDR transmission (National Instruments USRP cards, already available in DISI laboratories).
• Implementation and emulation of the closed-loop structure, with an SDN agent that decides, based on scenarios of technical interest in the non-terrestrial field, the transmission format and the management of the transceiver's reconfigurability (with issues of synchronization, compatibility, overhead, computational and energy sustainability, etc.).
The research project will exploit the synergies offered by the technical panel of the IEEE Aerospace and Electronic System Society (AESS): “Glue Technologies for Space Systems,” as well as interactions with partners involved in the follow-up of the above-mentioned PNRR projects, in particular with the ITA-NTN Proof-of-Concept development group concerning a drone equipped with a 5G payload, with the satellite networks laboratory of the Polytechnic University of Bari, and with the company PicoSats srl of Trieste.
[1] C. Sacchi et al., "A Unified Software-Defined Radio Framework for Flexible Waveform Design in Non-Terrestrial Networks," 2025 IEEE Aerospace Conference, Big Sky, MT, USA, 2025, pp. 1-20, doi: 10.1109/AERO63441.2025.11068553.
Contact: Claudio Sacchi claudio.sacchi@unitn.it
C1 - From behavioural streams to Spatio-temporal Knowledge Graphs
This fellowship will concentrate on the development of Linguistic resources will be used to provide Large Language Models with Common Sense Knowledge
Contact: Fausto Giunchiglia fausto.giunchiglia@unitn.it
C2 - Navigating the Societal Trajectory of AI: Socio-Technical Dynamics for Sustainable Organizational Impact and Individual Agency (Project: UE HE RIA - ELIAS – SEBE; G.A. 101120237; CUP E63C23000660006) (1 grant)
This project investigates the societal trajectory of Artificial Intelligence (AI), focusing on the socio-technical interface between automated decision-support systems and human agency in organizations. The research examines how AI tools are interpreted and operationalized, reshaping professional autonomy, cognitive processes, and accountability. It maps the friction between system deployments and lived experience, treating AI adoption as a complex organizational process. The study also addresses the “Evaluation Vacuum” in the European Research Area by analyzing how AI affects decision-making, service quality, and the sustainability of organizational practices. Using an interdisciplinary approach that bridges Innovation Studies, Sociology of Technology, and Computational Social Science, the project models the co-evolution of technical systems, social structures, and sustainable outcomes. The research seeks to understand how AI can produce accountable, long-term benefits for individuals and society, highlighting the interplay of technology, power dynamics, and sustainable organizational change.
Contact: Alberto Montresor alberto.montresor@unitn.it Niculae Sebe niculae.sebe@unitn.it
D1 - Artificial intelligence and data processing for planetary radars (Project: ASI JUICE RIME 2021 Bruzzone; CUP F65F21000950005. Project: ASI EnVision fase B1 Bruzzone; CUP F63C22000650005) (1 grant)
The research is related to data processing and machine learning for the analysis of data acquired by planetary radar sounders. Radar sounders operate from satellite platforms and acquire data related to the subsurface of planetary bodies that can results in groundbreaking science results.
The PhD will be developed in the framework of radars on board of two planetary missions of European Space Agency, i.e., the Sub-surface Radar Sounder (SRS) on board the EnVision mission to Venus of the European Space Agency (ESA) (for more information refer to https://www.esa.int/Science_Exploration/Space_Science/Envision) and the Radar for Icy Moon Exploration (RIME) on board of the JUpiter ICy moons Explorer (JUICE) of ESA (see https://www.esa.int/Science_Exploration/Space_Science/Juice).
The activity will be focused on the development of methodologies for the processing and the automatic analysis of the above-mentioned data for information extraction (e.g., classification, semantic segmentation, data inversion). Special emphasis will be given to methodologies that exploit the most recent developments in the framework of deep learning.
Research will be developed at the Remote Sensing Laboratory (https://rslab.disi.unitn.it/)
Contact: Lorenzo Bruzzone lorenzo.bruzzone@unitn.it
D2 - Artificial intelligence and deep learning for the analysis of satellite remote sensing images (Project: ESA HRLC_cci+ 2023 – Bruzzone; CUP E63C23002010006. Project: ESA - DISI - CC+ FASE 3 Bruzzone; CUP E63C25001200002) (1 grant)
The automatic analysis of image time series acquired by Earth Observation satellites is crucial for many different applications (e.g., precision farming, forestry, analysis of urban areas, monitoring of natural disasters). One of the most important applications is related to the monitoring of the land cover in relation to climate changes. In this context, even if artificial intelligence and machine learning have been widely used for the automatic analysis of remote sensing data, there are still many methodological challenges and application issues that should be addressed in the analysis of long time series of satellite data.
The research activities of this grant are related to the development of novel methodologies based on deep learning for the automatic analysis of time series of optical (multispectral) images acquired by Earth Observation satellites. The research will be focused on the problems of the automatic classification (semantic segmentation) and change detection in multispectral images acquired by different satellite systems. The activity will consider methodological research (for the development of novel methods) and the related application to real world scenarios.
Research will be developed at the Remote Sensing Laboratory (https://rslab.disi.unitn.it/) and will be linked to the High Resolution Land Cover project developed under the leadership of RSLab in the framework of the Climate Change Initiative of the European Space Agency (https://climate.esa.int/en/projects/high-resolution-land-cover/).
Contact: Lorenzo Bruzzone lorenzo.bruzzone@unitn.it
D3 / D4 - Development and clinical validation of numerical and experimental models for the study of ultrasound mechanisms of interaction with pulmonary tissue (Project: UE HE ERC LUMI Demi; GA n. 101229766; CUP E63C26000440006) (2 grants)
The research will be articulated in four layers, with the ultimate aim to generate new quantitative multi-frequency ultrasound methods for the diagnosis and monitoring of lung diseases.
The first layer revolves around the design and generation of numerical phantoms developed with increasing degree of complexity allowing the understanding of how different physical phenomena (scattering, absorption, bubble dynamics) play a role in characterizing ultrasound signals backscattered from lung tissue. The second layer concerns the fabrication and testing of novel and diverse physical models capable of mimicking lung properties under different pathological conditions to a level yet unseen. Phantoms will be made following two distinct and complementary approaches, i.e., clusters of air bubbles trapped in tissue mimicking gelatine and 3D printed phantoms. As for the numerical phantoms, experimental phantoms will be fundamental to improve our understanding of ultrasound wave’s penetration into and interaction with lung tissue. The third layer concerns the use of the first two layers output to generate in-silico and in-vitro lung ultrasound datasets with controllable ground truth. Numerical simulations will be performed employing state of the art ultrasound propagation simulators. Experimental data will be collected with advanced open-platforms for ultrasound research. Beyond the state of the art, limited to 2D datasets, high frame rate and 3D datasets will be generated. The entire dataset will be made publicly available. The fourth layer is then focused on clinical translation by means of investigating the robustness of the in-silico and in-vitro results on clinical data.
Research will be conducted in a highly international environment.
Contact: Libertario Demi libertario.demi@unitn.it
Fondazione Bruno Kessler (FBK)
A4 - Artificial Intelligence for Tiny, Connected Devices: Enabling Learning and Inference on Resource-Limited Networked Embedded Systems (1 grant)
The Internet of Things (IoT) paradigm is driving a massive increase in the generation of multimodal data on tiny, resource-constrained devices at the far edge of the computing infrastructure. Given the challenges of limited computation, energy constraints, and communication bottlenecks, there is a growing need to process data locally while ensuring efficient and scalable artificial intelligence at the edge. TinyML and Edge AI have demonstrated the feasibility of embedding machine learning models on such devices. Still, many challenges are ahead. Expanding their impact in real-world deployments requires addressing the heterogeneity of hardware, data, and resource availability in distributed scenarios.
This research will explore novel approaches for enabling AI at the edge, focusing on one or all of the following aspects:
i) hardware-aware scaling, model compression, and novel approaches for diverse low-power edge devices in distributed and collaborative IoT scenarios;
ii) strategies for efficient and adaptive learning on-device or across a network of heterogeneous nodes while minimizing energy consumption and bandwidth usage;
iii) investigating how explainability and robustness can be maintained in compressed models deployed at the far edge, ensuring trustworthiness and reliability in real-world applications
This interdisciplinary research at the intersection of Artificial Intelligence, Embedded Systems, Distributed Computing, and Low-power Hardware and protocols will be tailored to the candidate's profile and interests, contributing to the development of innovative solutions for real-world challenges.
Contact: Elisabetta Farella efarella@fbk.eu
A5 - Controlling the behavior of collective of LLMs (1 grant)
The successful PhD candidate will do research on Large Language and Reasoning Models (LLMs/LRMs), with a focus on interpretability and steerability of their behavior. A particular focus of the thesis will be on the design and monitoring of collective systems populated by AI generative agents, and on improving their theory of mind, lifelong learning and social interaction abilities.
Contact: Bruno Lepri lepri@fbk.eu
A6 - Human-centred Evaluation Frameworks for Multilingual Technologies (1 grant)
How we measure and define performance shapes the systems we build. Current Natural Language Processing (NLP) benchmarks often prioritize leaderboard scores over practical utility, failing to capture how models behave in real-world, socially situated contexts. This PhD project treats evaluation methodology as a research problem in its own right, advancing both the conceptual foundations and computational tools for assessing NLP systems in ways that are reliable, valid, and human-centred. Application domains include multilingual settings, such as machine translation, and emerging agentic and interactive multimodal NLP systems involving human-AI collaboration present frontier evaluation challenges. The ultimate goal is to develop evaluation frameworks that capture not only overall system performance, but also real-world utility, fairness, and responsiveness to the needs of diverse stakeholders.
Contact: Matteo Negri negri@fbk.eu
A7 - Multimedia Personalization with Multimodal Large Language Models (1 grant)
The rise of multimodal large language models (MLLMs) is transforming language, speech, and vision technologies, enabling unprecedented capabilities in translation, summarization, and, in general, content generation. This PhD project will focus on personalization, exploring how models can adapt to individual users’ preferences, context, and communication style. Research directions include adaptive speech-to-speech translation, context-aware description generation, text simplification, and modeling users’ preferences. By integrating these approaches into MLLMs, the project aims to create more natural, context-sensitive, and user-centric multilingual experiences, pushing the boundaries of how AI can serve people on an individual level.
Contact: Matteo Negri negri@fbk.eu
B4 - Artificial Intelligence for Tiny, Connected Devices: Enabling Learning and Inference on Resource-Limited Networked Embedded Systems (1 grant)
The Internet of Things (IoT) paradigm is driving a massive increase in the generation of multimodal data on tiny, resource-constrained devices at the far edge of the computing infrastructure. Given the challenges of limited computation, energy constraints, and communication bottlenecks, there is a growing need to process data locally while ensuring efficient and scalable artificial intelligence at the edge. TinyML and Edge AI have demonstrated the feasibility of embedding machine learning models on such devices. Still, many challenges are ahead. Expanding their impact in real-world deployments requires addressing the heterogeneity of hardware, data, and resource availability in distributed scenarios. This research will explore novel approaches for enabling AI at the edge, focusing on one or all of the following aspects: i) hardware-aware scaling, model compression, and novel approaches for diverse low-power edge devices in distributed and collaborative IoT scenarios; ii) strategies for efficient and adaptive learning on-device or across a network of heterogeneous nodes while minimizing energy consumption and bandwidth usage; iii) investigating how explainability and robustness can be maintained in compressed models deployed at the far edge, ensuring trustworthiness and reliability in real-world applications This interdisciplinary research at the intersection of Artificial Intelligence, Embedded Systems, Distributed Computing, and Low-power Hardware and protocols will be tailored to the candidate's profile and interests, contributing to the development of innovative solutions for real-world challenges.
Contact: Elisabetta Farella efarella@fbk.eu
B5 - Dependable and Energy-Aware Cybersecurity for Cyber-Physical Production Systems (1 grant)
Modern productive environments increasingly rely on complex cyber-physical systems (CPS) to manage operational processes. While this digitalisation enables higher efficiency and flexibility, it also significantly expands the cyber-attack surface of production systems, exposing CPS to cyber and cyber-physical threats, including side-channel attacks that exploit unintended information leakage from control logic, timing behaviour, or energy consumption patterns, which can compromise safety, availability, and energy efficiency.
This PhD project focuses on the development of advanced cybersecurity solutions for anomaly detection in energy-aware cyber-physical production systems. The research will investigate robust and dependable AI-based techniques capable of detecting both known and previously unseen cyber and cyber-physical attacks, including side-channel-based threats, that affect operational integrity and energy management processes. Particular attention will be given to the fusion of multi-modal data sources, including industrial network traffic, control signals, process variables, and energy-related measurements characteristic of CPS, which may reveal anomalous or covert behaviours.
The project will explore adaptive and real-time anomaly detection methods for CPS, as well as explainable AI (XAI) techniques to support human operators in understanding security incidents and assessing their impact on energy usage, system reliability, and production continuity.
Contact: Marco Zambianco mzambianco@fbk.eu
B6 - Secure AI-Focused Programming and Regulatory Software Compliance (1 grant)
The global workforce is facing critical staff shortages due to aging populations and declining birth rates. This demographic shift threatens productivity, especially in fields like computer science and cybersecurity, where demand for skilled professionals is rising alongside system complexity and regulatory requirements (Piras et al. 2020). As ICT companies expand, the need for software and security engineers grows, yet programming remains a time-consuming and intricate task.
To sustain productivity, automation and optimisation are essential. These strategies streamline the software engineering cycle by enabling higher abstraction levels, reducing time and resource demands, and improving code security. AI offers promising support, such as tools that generate code (Vaithilingam et al. 2022). However, current AI-generated code often contains errors and lacks scalability, requiring manual debugging and refinement. While AI can produce valid snippets, it struggles to create complete, secure systems at scale.
A promising solution is decomposing complex systems into smaller sub-problems, allowing AI to generate components more effectively. Goal Modeling (GM) techniques are well-established for this purpose, enabling hierarchical decomposition of requirements (Horkoff et al. 2019). GM and visual programming help developers work at higher abstraction levels, potentially accelerating design.
We hypothesize that combining GM with AI can democratize software development, enabling broader participation beyond traditional programmers. Integrating this approach with Symbolic AI, Neuro-Symbolic AI, Agentic AI, Neural Networks for code vulnerability detection (Senanayake et al. 2024), SBOM tools, prompt vulnerability detectors, and static/dynamic analysis tools could enhance code security and regulatory compliance (Piras et al. 2020). These synergies could empower diverse professionals to contribute to software engineering, helping address workforce shortages while building secure, high-quality systems.
This research proposes designing a secure, AI-enhanced goal-modeling framework to support rapid software development with reduced skill requirements. By “secure,” we refer to integrating software security measures and ensuring compliance with regulations like the EU AI Act, NIS2 and GDPR (Piras et al. 2020). The research may focus on conceptualizing and prototyping this framework, identifying key components, and evaluating it in realistic or industrial settings, potentially in collaboration with industry partners.
Contact: Luca Piras l.piras@fbk.eu
B7 - Supporting AI Act Compliance via an Intelligent Holistic Environment (1 grant)
Artificial Intelligence (AI), particularly with the introduction of LLMs, is transforming our society while raising significant regulatory concerns. In response, the EU introduced the EU AI Act.
The AI Act is a complex regulation that classifies AI systems according to risk levels and introduces complex transparency and compliance requirements as well as high-level indication of potential enforcement mechanisms. Organisations must ensure compliance to avoid legal sanctions or system suspension. Given the complexity of the AI Act, effective compliance requires collaboration among AI experts, IT professionals, legal specialists, and other stakeholders involved in the design, deployment and use of AI systems.
Challenges, costs and effort required to comply with the AI Act are likely to resemble those experienced with the introduction of GDPR, which imposed a demanding privacy-by-design approach (Tsohou et al. 2020; Piras et al. 2020). GDPR compliance required staff training, process redesign and collaboration among diverse experts to adapt systems and practices. Consequently, new tools and methods were developed to support organisations (Bhalavat et al. 2024; Piras et al. 2020). Similar support mechanisms will be necessary to reduce effort, cost, and time needed for AI Act compliance (Kulkarni et al. 2021).
Current approaches supporting AI Act compliance mainly rely on specific tools and sandboxes, designed to detect specific AI biases and risks, while valuable, such solutions are often limited in scope, do not account for collaboration among heterogeneous professionals, and rarely address the entire AI lifecycle. However, AI systems require continuous monitoring, especially when updated or retrained, as new biases or risks may emerge and affect compliance. What is still missing is an integrated environment that supports organisations holistically, considering regulatory, technical and organisational aspects, enabling collaboration among diverse stakeholders, and continuously monitoring AI systems. Such an environment, potentially using LLMs, could assist analysts in promoting compliance and anticipating potential non-compliance, for instance through predictive mechanisms such as a digital twin.
The candidate of this research may focus on the design of the concept and prototype of such research environment for supporting organisations towards AI Act compliance, by identifying some of the most important aspects contributing to the compliance, creating the environment for supporting the collaboration of different professional roles, and evaluating it in realistic/real settings, potentially from some of our industry partners, with the use of critical and relevant scenarios.
AI Act is the currently most interesting regulation to consider, however this research can consider and explore compliance with other regulations (e.g., NIS2, EHDS, CRA, DORA), and potentially cross-regulatory compliance.
Contact: Luca Piras l.piras@fbk.eu
C3 - Failure Propagation Analysis for Safety Assessment of Complex Systems (1 grant)
The design process of complex systems must guarantee not only the functional correctness of the implemented system, but also its safety, dependability, and resilience with respect to run-time faults. To this aim, complex systems implement mechanisms to timely detect components' faults and to isolate them, before they can propagate and cause system failures. Hence, the design process must characterize the likelihood and severity of faults, identify the set of possible hazards and failure conditions, mitigate possible consequences, and assess the effectiveness of the adopted mitigation measures.
Model-Based Safety Analysis (MBSA) is listed as an acceptable and recommended means of compliance to perform safety assessment in the latest issue of SAE ARP4761A, specifically for analyzing failure propagation. MBSA is based on the adoption of a formal, mathematical model of the system and on a tool-supported methodology to assist the generation of safety artifacts. State-of-the-art tools for MBSA implement functionalities to generate Minimal Cut Sets (MCS) from a fault propagation model and a Top-Level Event (TLE) [IMBSA25, LPNMR22, CAV21]; perform automated fault injection into a behavioral design model to generate the corresponding safety model [FAOC21, TACAS16]; generate Minimal Cut Sets from a fully behavioral dynamical model and a TLE [FAOC21, TACAS16, CAV15a, SCP15]; perform various kind of validation of fault propagation models against behavioral models [IJCAI16, AAAI16, AAAI15].
The objective of this study is to advance the state-of-the-art in failure propagation analysis and safety assessment of complex systems. In particular, it will investigate extensions of existing formalisms to deal with aspects such as the timing of fault propagation, the characterization of transient and sporadic faults, and the analysis of the effectiveness of fault mitigation measures in presence of complex fault patterns. Moreover, this study will investigate the use of fault propagation models for the design of fault detection, isolation and recovery (FDIR) components. To this aim, fault propagation models will be extended with observability information and used to solve problems such as anomaly detection, diagnosis, root-cause analysis, and prognosis. Finally, this study will aim to bridge the gap between fault propagation models and fully behavioral system models used for the design and safety assessment of complex systems.
Contact: Marco Bozzano bozzano@fbk.eu
C4 - Methodologies for automated testing of complex, configurable systems (1 grant)
The increasing complexity of software systems necessitates the development of methodologies and tools for designing and testing systems exhibiting variability in the domain of potential functional configurations, in conjunction with diverse release architectures and a growing level of run-time autonomy.
The principal objective of this PhD is to explore innovative strategies that facilitate the automated generation of test scenarios and the construction of corresponding testing oracles for such systems. This will be achieved by employing agentic artificial intelligence, model-based techniques, and optimisation methodologies.
Contact: Angelo Susi susi@fbk.eu
C5 - Modeling and Simulation of Urban Digital Twins (1 grant)
An Urban Digital Twin is a dynamic digital model of a city, fed by data collected from the city itself, and capable of faithfully reproducing the city behaviour through the use of advanced modelling, data science, and AI techniques. One of the key foreseen applications of Urban Digital Twins is to predict the evolution of the city, including the effects and impacts of external changes (eg, climate change) and internal processes (eg, urban transition policies and incentives).
A challenge for the development of Urban Digital Twins is that cities are systems-of-systems, with complex interactions between physical, organisational, and social dimensions, as well as between the physical world and the digital world. Novel modeling and simulation techniques are necessary to develop Urban Digital Twins able to manage this complexity and produce reliable predictions.
The candidate will be requested to contribute to this research area, and in particular to work on advanced modeling and simulation frameworks for Urban Digital Twins. The expected contribution of the candidate is twofold: first, developing a novel framework of new analytical methods for modeling and simulation in the Urban Digital Twin; second, assessing the validation of the proposed framework on real-world scenarios concerning the adoption of Digital Twin by Italian cities.
Besides the requirements established by the rules of the IECS school, preferential characteristics for candidates for this scholarship are:
- Master degree in Computer/Data Science, Mathematics, Physics, Electrical Engineering, Communication Engineering, or equivalents;
- Knowledge in artificial intelligence, statistical and machine learning, complex systems, agent-based modeling and simulation.
Contact: Marco Pistore pistore@fbk.eu
C6 - Symbolic Model Checking techniques for embedded systems (1 grant)
Techniques based on formal methods for the verification and validation of embedded and safety-critical systems are becoming increasingly important, due to the growing complexity and importance of such systems in every aspect of modern society. Despite the major progress seen in the last twenty years, however, the application of formal methods remains a challenge in practice, due to factors such as poor scalability, lack of automation, the interplay between computation and physical aspects, or the increasing complexity of the software and its configurations.
This project will investigate novel techniques for the application of formal methods to the design, verification, and validation of embedded systems, with particular emphasis on safety-critical application domains such as railways, automotive, avionics, and aerospace. A particular attention will be devoted to improving the scalability and degree of
automation of formal methods techniques, with a specific focus on symbolic model checking methods using satisfiability and satisfiability modulo theories solvers as symbolic reasoning engines. Importantly, in addition to researching novel theoretical results, a significant part of the project activities will be devoted to the implementation of the techniques in state-of-the-art verification tools developed at FBK and their application to real-world problems in collaboration with our industrial partners.
Contact: Alberto Griggio griggio@fbk.eu
D5 - Artificial intelligence and machine learning for radar sounder data processing (1 grant)
Within the framework of the European Space Agency’s Jupiter Icy Moons Explorer (JUICE) and EnVision mission to the Jovian system and Venus respectively, we seek highly motivated candidates to develop innovative methodologies for radar sounder data processing, with a strong emphasis on artificial intelligence and deep learning. The research activity will focus on the design and implementation of advanced models for data enhancement, denoising, semantic segmentation, content-based retrieval, target detection, multitemporal analysis, and radar mapping, aiming at improving the information extraction and the of icy planetary bodies subsurface by radar sounder data processing. The selected candidate will be expected to develop novel machine learning methods, including deep learning architectures, self-supervised and unsupervised approaches, physics-informed neural networks, transformer-based models, and/or quantum-inspired learning techniques, capable of integrating electromagnetic propagation physics with data-driven strategies, while addressing key challenges such as noise, clutter, signal attenuation, and limited availability of labelled data. The outcomes of this research will significantly advance understanding of planetary subsurface structures and their geological and climatic evolution, and will develop transferable methodologies applicable to Earth observation.
Besides the requirements established by the rules of the ICT school, the following characteristics are preferred for candidates for this scholarship:
- master's degree in Electrical Engineering, Communication Engineering, Computer/Data Science, Mathematics, or equivalent;
- knowledge in artificial intelligence, image/signal processing, remote sensing, and radar remote sensing.
Contact: Francesca Bovolo bovolo@fbk.e
D6 - Artificial intelligence for remote sensing image time series analysis (1 grant)
The recent Earth Observation missions like (ESA Copernicus - Sentinels, ASI PRISMA and COSMO-SkyMed, and future IRIDE constellation) make available databases of long, dense and worldwide image time series. The data have complex spatio-spectro-temporal behaviors and variability, and they show irregularities and misalignments, yet they allow for a wide range of downstream tasks.
Candidates will be requested to develop novel methodologies within the artificial intelligence framework for effectively and efficiently process image time series for semantic segmentation, target detection and change detection along and across multiannual series of data. Methodologies like foundational models, machine learning, deep learning, multitask learning, enforcement learning, explainable AI, etc. will be considered.
Besides the requirements established by the rules of the ICT school, preferential characteristics for candidates for this scholarship are:
- master degree in Electrical Engineering, Communication Engineering, Computer/Data Science, Mathematics or equivalents;
- background in artificial intelligence, image/signal processing, remote sensing, passive/active sensors.
Contact: Francesca Bovolo bovolo@fbk.eu
D7 - Generalization in Vision-Language-Action models (1 grant)
Despite the success of large-scale pre-training, Vision-Language-Action (VLA) models often exhibit limited generalization when deployed in novel settings. These models can fail when encountering shifts in visual domains, diverse environmental contexts, or variations in robotic embodiments. Such limitations hinder the practical application of foundation models in open-world robotics. This doctoral project aims to investigate the factors that restrict VLA generalization and develop strategies to improve their adaptability. The research will explore techniques such as cross-domain alignment, robust representation learning, and embodiment-agnostic architectures. The goal is to create robot learning frameworks that can reliably transfer knowledge across different tasks and physical platforms, ensuring consistent performance in unseen real-world scenarios.
Contact: Davide Boscaini dboscaini@fbk.eu
D8 - Generative vision models for robot learning (1 grant)
Modern robotics is at a pivotal point. While traditional learning-based methods have shown success in navigation and manipulation, they remain hampered by high data collection costs and poor out-of-the-box transferability. This doctoral project aims to bridge the gap between Generative AI and Physical Embodiment. The goal is to investigate closed-loop robotic systems capable of interacting with complex, real-world environments by integrating vision, language, and proprioception into a unified generative framework.
Contact: Guofeng Mei gmei@fbk.eu