Federico De Lorenzis

Several collaborations are currently in place with the Piedmont Region Civil Protection and Forest Firefighting Unit, which are actively involved in emergency management and first responder training. In this context, a number of serious games and interactive experiences leveraging Virtual Reality (VR), Augmented Reality (AR), and Mixed Reality (MR) technologies have already been developed to support the education and training of firefighters, volunteers, students, and decision makers.
Multiple thesis opportunities are available and will be developed mainly at Politecnico di Torino in collaboration with the above-mentioned institutions.
The proposed activities build upon an existing MR-based training ecosystem previously developed through several research and thesis projects, including the “Augmented Sand Table for Forest Firefighting”, currently supporting wildfire simulation scenarios based on the Cell2Fire framework. Current research activities focus on extending this ecosystem toward more immersive, scalable, and interoperable XR training solutions for emergency-response scenarios, with particular emphasis on wildfire management and hydrogeological risk mitigation. The activities may involve the use of Unity, XR headsets, simulation middleware, networking solutions, and advanced interaction techniques.
The objective of the first thesis is to extend and enhance existing immersive XR training applications for first responders operating in complex emergency scenarios, including wildfire suppression activities involving both ground and aerial operations. The work may address the integration of more realistic environmental simulations, interactive tools, and collaborative functionalities aimed at improving training effectiveness and operational realism. Particular attention may be devoted to fire and smoke simulation, immersive interaction techniques, real-time performance optimization, and interoperability with external simulation systems.
The goal of the second thesis is to evolve the existing simulation ecosystem through the development of advanced authoring tools and workflow solutions aimed at simplifying and accelerating the generation, configuration, and customization of XR training scenarios and case studies. The work may address procedural and data-driven scenario generation techniques, terrain editing solutions, and interoperability mechanisms for integrating heterogeneous simulation frameworks into scalable and reusable XR training ecosystems. A particular focus may be placed on extending the current Cell2Fire-based simulation pipeline toward additional emergency-management scenarios, including the integration of simulation systems addressing hydrogeological risks such as floods and terrain evolution.
The third thesis focuses on the evolution of previously developed XR applications into a unified multi-layer and multi-user ecosystem for coordinated emergency management. The resulting platform should support supervisors and decision makers in monitoring, coordinating, and evaluating the actions of distributed teams operating within interconnected simulation scenarios. The work may involve the synchronization of heterogeneous simulators, the implementation of collaborative XR interaction paradigms, the development of monitoring and analytics solutions, and the exploration of AI-assisted approaches for scenario management and operational analysis. The thesis may also explore the integration of immersive simulation modules for aerial wildfire suppression operations involving helicopters and Canadair aircraft within larger coordinated simulation environments.
References:
– https://www.knowledge-share.eu/en/software/tavolo-di-sabbia-aumentato-per-la-lotta-agli-incendi-boschivi
– https://ieeexplore.ieee.org/document/10902197
– https://ieeexplore.ieee.org/abstract/document/8576892

Virtual Reality (VR) has proven effective in simulating procedural tasks and structured training scenarios, particularly in domains where experiential learning is critical. However, current VR learning experiences are often limited in their ability to support dynamic problem-solving, especially in situations where learners are required to devise tools, strategies, or solutions in real time. This lack of adaptability reduces learner agency and constrains the exploration of alternative approaches within immersive environments.
This thesis aims to investigate the use of Generative Artificial Intelligence to enhance learner agency in VR-based training systems. The objective is to enable users to interactively solve training scenarios by leveraging AI to generate context-aware suggestions, including tools, hypotheses, and procedural steps. In addition, the work will explore the integration of Reinforcement Learning techniques to adapt the behavior of the AI assistant over time, optimizing when and how support is provided based on user interactions and learning progress.
The student will design and develop prototype systems in which users can invoke AI assistance during VR training sessions. The approach will initially focus on unconstrained environments, where learners can freely explore AI-generated solutions in a sandbox setting. Subsequently, domain-specific constraints will be introduced to guide the generation process toward feasible and coherent solutions aligned with the procedural logic of the task. Reinforcement Learning strategies will be investigated to model adaptive assistance policies, allowing the system to personalize feedback, regulate the level of guidance, and improve its effectiveness through interaction. Particular attention will be devoted to balancing flexibility and control, as well as to the definition of meaningful reward signals based on user behavior, task completion, and learning outcomes.
The developed systems will be evaluated through qualitative and quantitative methods, considering aspects such as user engagement, perceived agency, solution diversity, coherence with domain knowledge, and overall usability. The expected outcome is an enhancement of learner creativity and autonomy through AI-assisted problem solving, along with the development of adaptive learning pathways tailored to user behavior. This work may also contribute to the definition of intelligent, responsive training environments capable of evolving in real time through the combined use of Generative AI and Reinforcement Learning.
References
[1] S. Doroudi, V. Aleven, and E. Brunskill, “Where’s the Reward? A Review of Reinforcement Learning for Instructional Sequencing,” in International Journal of Artificial Intelligence in Education 29, 2019, pp. 568-620. doi: 10.1007/s40593-019-00187-x.
[2] S. Yao et al. “ReAct: Synergizing Reasoning and Acting in Language Models,” arXiv:1907.00456, 2019, pp. 1-16, doi: 10.48550/arXiv.1907.00456.
[2] N. Jaques et al. “Way Off-Policy Batch Deep Reinforcement Learning of Implicit Human Preferences in Dialog,” arXiv:2210.03629, 2023, pp. 1-33, doi: 10.48550/arXiv.2210.03629.

Recent advances in Generative Artificial Intelligence are opening new possibilities for the design and development of immersive Virtual Reality (VR) learning experiences. Traditional VR authoring pipelines rely heavily on manual processes, including 3D modelling, scripting, and scenario design, which limit scalability, adaptability, and responsiveness to evolving training needs. At the same time, generative AI techniques are increasingly being integrated into VR systems to automate the creation of environments, assets, and interactions, enabling more dynamic and personalized experiences.
This thesis aims to investigate the integration of Generative AI into VR-based training systems, with the objective of automating the creation of virtual environments, interactive objects, and narrative logic. The work will explore how textual or semantic inputs can be leveraged to generate coherent VR experiences by combining generative models with procedural content generation techniques. In particular, hybrid approaches will be considered, where text prompting guides the generation process while procedural systems enforce constraints, consistency, and scalability of the resulting environments.
The student will design and develop prototype systems supporting semi-automatic or fully automatic authoring of VR training scenarios. This includes the integration of large language models for scenario generation, generative models for 3D asset creation, and procedural methods for assembling environments and interactions. Particular attention will be devoted to the interaction between generative and procedural components, the definition of intermediate representations, and the design of prompting strategies for controllable and reliable generation.
The developed systems will be evaluated through both qualitative and quantitative analyses, considering factors such as generation quality, coherence of the resulting experiences, adaptability to different training contexts, and usability. The expected outcome is a reduction in development time and cost for immersive training applications, along with contributions toward a structured framework for integrating Generative AI and procedural techniques into VR authoring pipelines, supporting scalable and adaptive training systems.
References:
[1] Y. Song, K. Wu, and J. Ding, “Developing an immersive game-based learning platform with generative artificial intelligence and virtual reality technologies – “LearningverseVR,” in Computers & Education: X Reality 4, 2024, pp. 1-7. doi: 10.1016/j.cexr.2024.100069.
[2] X. Mao, “Procedural Content Generation via Generative Artificial Intelligence,” arXiv:2407.09013, 2024, pp. 1-15, doi: 10.48550/arXiv.2407.09013.
[3] F. Rahimi, A. Sadeghi-Niaraki, and S. Choi “Generative AI Meets Virtual Reality: A Comprehensive Survey on Applications, Challenges, and Future Direction,” in IEEE Access, vol. 13, pp. 94893-94909, 2025, doi: 10.1109/ACCESS.2025.3574779.

Problem:
VR-based learning experiences, especially those centered on procedural aspects, demand repeated practice sessions to be effective. However, these experiences often suffer from limited variety, which can hinder educational outcomes due to the principle of “encoding specificity”.
Objective:
Studying and adapting core mechanics from the rogue-lite genre (encompassing the procedural generation of scenarios and retention of progress across game sessions) to create engaging, effective, and coherent VR learning environments and training systems.
Approach:
Select an appropriate learning procedure; design and develop a VR application integrating rogue-lite mechanics to improve procedural skills. Validate the tool and compare it to a static learning scenario
Expected Impact:
A new perspective on how to design and implement effective VR learning experiences

Problem:
Students often struggle to engage deeply with complex topics, and most study tools do not support the cognitive benefits of learning-by-teaching. Without a system that lets learners explain and refine knowledge for a “novice,” they miss a powerful mechanism for strengthening understanding.
Objective:
Study and develop a platform where students “teach” an agent that simulates progressive learning. Using structured memory and adaptive dialogue, the system will tailor future interactions based on what the student has explained.
Approach:
Select a suitable topic; design and develop an agent (e.g., a chatbot) that builds knowledge from student input through adaptive memory. Validate the tool and compare it to a static, non-adaptive chatbot.
Expected Impact:
A more engaging learning environment that leverages the learning-by-teaching paradigm to improve comprehension, metacognition, and student motivation.

Problem:
Traditional Virtual Tours often lack real-time responsiveness to user behavior. Virtual guides often fail to ensure group cohesion or tailor content dynamically, leading to disengagement and fragmented experiences.
Objective:
To develop AI-powered tools capable of tracking user behavior, maintaining group coherence, and adapting content delivery to individual and collective engagement patterns in digital museum and showroom tours.
Approach: The system will integrate real-time user tracking, behavioral modeling, either to adapt content presentation or to offer valuable hint and suggestion to the designated guide. Machine learning algorithms will infer user intent and attention, enabling the guide to adjust pacing, highlight relevant artifacts, and ensure no participant is left behind.
Expected Impact:
This research is expected to improve user engagement and retention in digital cultural experiences by enabling intelligent, behavior-aware guidance. It will support more inclusive and personalized tours, where virtual curators
adapt dynamically to group cohesion and individual interests. The outcomes may inform future design strategies for scalable, AI-enhanced museum and showroom applications.