The computer graphics and computer vision techniques are essential for the analysis and the visual representation of key information regarding the state of production processes. In processes that envisage a human intervention, e.g., in scenarios based on the use of collaborative robots (or cobots), they can support the identification and managements of defects and issues in general; they can additionally contribute at increasing the safety of the workplace, through the study/the simulation of ergonomics aspects, the delivery of training contents, the monitoring/prediction of human actions in relation to the machine ones.

In these fields, several thesis are available, framed in the context of the MANAGE 5.0 project and developed in collaboration with Stellantis and other companies, with the support by Ministry of Enterprises and Made in Italy (MISE) and the National Recovery and Resilience Plan (PNRR).

In particular, the MANAGE 5.0 project will focus on the definition of intelligent approaches for the creation of digital twins of production lines involving cobots and of their operators, and on the use of immersive technologies (virtual, augmented and mixed reality) for the optimization of human-machine collaboration, onsite and at distance via telepresence solutions.

The computer graphics and computer vision techniques are essential for the analysis and the visual representation of key information regarding the state of production processes. In processes that envisage a human intervention, e.g., in scenarios based on the use of collaborative robots (or cobots), they can support the identification and managements of defects and issues in general; they can additionally ...

Several thesis proposals are available in the context of the "Centro Nazionale Mobilità Sostenibile" ("National Center for Sustainable Mobility"), in particular of the project "Air mobility" funded by the National Recovery and Resilience Plan (PNRR). These proposal will concern the design and development of virtual simulations and human-computer interfaces for advanced air mobility, and could be developed in collaboration with the other project partners.

One of these proposals will concern the creation of a simulation environment for flying vehicles. Starting from existing software, an immersive environment capable to simulate, both visually and physically, the cabin/cockpit of a Vertical Take-Off and Landing (VTOL) aircraft (manned or un-manned) will be created. The physical simulation will leverage a motion platform available at VR@POLITO. Typical tasks for VTOL aircrafts like passenger transport, emergency medical services, or search and rescue (SAR) will be considered.

Another proposal strictly related to the previous one will concern the design and evaluation of interfaces capable to enhance the situational awareness of pilots of manned aircrafts. In particular, since the two main human-vehicle communication channels, i.e., the visual and aural ones, are typically overloaded, the aim is to investigate how to leverage the haptic channel. In particular, by using a commercial haptic suit, the plan is to investigate the effectiveness of a multi-sensory stimulation combining the above channels with the haptic one by means of full-body touch stimuli.

Another proposal will concern the use of spatial audio in immersive VR simulations to investigate the acoustic experience in flying vehicles. Though a collaboration with the DIMEAS department, acoustic simulations of the the vehicle's mechanical structure and motors will be generated. By means of a VR simulation, users will be immersed a synthetic environment and will be allowed to evaluate different configurations based on their acoustic impact.

Several thesis proposals are available in the context of the "Centro Nazionale Mobilità Sostenibile" ("National Center for Sustainable Mobility"), in particular of the project "Air mobility" funded by the National Recovery and Resilience Plan (PNRR). These proposal will concern the design and development of virtual simulations and human-computer interfaces for advanced air mobility, and could be developed in ...

The recognition of the vehicle’s owner or the person authorized to operate a given vehicle will be a challenge for future intelligent vehicle applications. Furthermore, the person to be recognized stands outside the car, in different and complex conditions (recognition in the “wild”): highly variable lighting, presence of other subjects, etc. Apart from the difficulty and complexity of the recognition algorithm, one of the fundamental problems becomes the data needed to train required AI models, which cannot be available yet.

In this case, so-called synthetic data need to be employed. Synthetic data, generated through some simulation environments, can mimic operational conditions, by also enabling to keep edge cases into account. In fact, real-world datasets often contain imbalances, because edge cases, which do not happen frequently in real life, are not sufficiently represented. Finally, with synthetic data, other issues, e.g., related to privacy and GDPR compliance, can be easily overcome.

The focus of this thesis, developed in collaboration with Centro Ricerche Fiat (CRF) and Stellantis, will be the use of computer graphics simulation for the the generation of synthetic data that could be later used to train AI algorithms for the recognition of vehicle owners that are robust to variations in terms of, e.g., illumination, presence of other human and non-human subjects, etc.

The recognition of the vehicle’s owner or the person authorized to operate a given vehicle will be a challenge for future intelligent vehicle applications. Furthermore, the person to be recognized stands outside the car, in different and complex conditions (recognition in the “wild”): highly variable lighting, presence of other subjects, etc. Apart from the difficulty and complexity of the recognition ...

Virtual humans (or digital humans) are simulations of human beings on computers, which are considered today a kind of commodity in movies, video games and the entertainment domain in general. In fact, virtual humans are getting commonplace in various industries and domains, from medicine, to film, fashion, education, automobile, telecommunications, etc. Depending on the application, various level of realisms might be needed.

A medical application requires an exact simulation of specific internal organs or of human emotional states; an education or a cultural heritage application requires effective communication/storytelling and interaction abilities to effectively deliver the intended contents and manage dialogues; the film industry requires extremely pleasant appearances, as well as natural movements and expressions; ergonomic studies, like clothing applications, require faithful body proportions, realistic locomotion, etc.; the metaverse require social behavior to be replicated in a credible way via easy-to-use interfaces.

The research domain of virtual humans is concerned with their representation, movement and behavior, and is addressing related challenges through a multi-disciplinary approach that encompasses, among others, computer graphics, computer animation, computer vision and artificial intelligence.

The aim of this thesis is to explore possible methods and tools for the creation and the control of intelligent virtual humans that can be used in different projects currently ongoing at the VR@POLITO lab.

In particular, activities could concern:

- the creation of digital avatars for virtual visits being created for cultural institutions in Piemonte and in Italy (in collaboration with DAD and DISEG departments);
- the development of empathic avatars of patient and doctors to support training in the medical domain in collaboration with OntarioTech university (Ontario, Canada);
- the implementation of conversational agents that can be used in customer service-oriented applications;
- the comparison of consumer technologies that can be exploited to enable a smooth transfer of the user's movements (body posture and gait, as well as facial expression) onto his or her virtual avatar.

Depending on the domain, the thesis can be carried out either in collaboration with a company, as a thesis at Politecnico di Torino or even as a thesis abroad in a foreign university.

Virtual humans (or digital humans) are simulations of human beings on computers, which are considered today a kind of commodity in movies, video games and the entertainment domain in general. In fact, virtual humans are getting commonplace in various industries and domains, from medicine, to film, fashion, education, automobile, telecommunications, etc. Depending on the application, various level of realisms ...

Autonomous vehicles are routinely advertized as the world-changing invention that will greatly impact people lives by, on the one hand, improving safety, reducing fuel consumption and decreasing urban infra-structure strains, and, on the other hand, freeing up huge amounts of time for their occupants, who will be allowed to read, consume, work, play, etc. Despite the enormous business opportunities, autonomous vehicles also come with some key issues.

One of the major issues is related to the fact that, in oder to use these technology, people will first have to trust it. Besides education and experience, a factor that is regarded as the possible keystone for constructing trust towards autonomous vehicles is represented by HMI, i.e., by the paradigms chosen to communicate with the users (the vehicle occupants, but also the other road users) and to make them interact with it.

At present, since just a few vehicles are actually available for real-world experiments (and not at the highest levels of autonomy envisaged by SAE, yet) the only way to design and validate approaches for possibly coping with issues like those above is through simulation.

The VR@POLITO has developed an immersive VR driving simulation platform that has been used, e.g.,

- to explore in-vehicle interfaces based on AR Head Up Displays for enhancing drivers'/passengers' trust;
- to study pedestrians' behavior at crosswalks with incoming vehicles mounting different types of signaling interfaces;
- to investigate the effectiveness of multi-modal interfaces in urban, suburban and highway drives with autonomous vehicles.

Several thesis works are available in this field, which can address the open problems from different perspectives, possibly in collaboration with a company.

Suggested readings:

Building Trust in Autonomous Vehicles: Role of Virtual Reality Driving Simulators in HMI Design
https://arxiv.org/abs/2007.13371

Comparing State-of-the-Art and Emerging Augmented Reality Interfaces for Autonomous Vehicle-to-Pedestrian Communication
https://arxiv.org/abs/2102.02783

Autonomous vehicles are routinely advertized as the world-changing invention that will greatly impact people lives by, on the one hand, improving safety, reducing fuel consumption and decreasing urban infra-structure strains, and, on the other hand, freeing up huge amounts of time for their occupants, who will be allowed to read, consume, work, play, etc. Despite the enormous business opportunities, autonomous ...

Several collaborations are in place with various institutions (Italian Airforce, Piedmont Region Civil Protection and Forest Firefighting Unit, Frejus Tunnel Authority, Fire and Rescue Department of the Savoy Region, France, etc.) involved in emergency management. In this context, a number of serious games and interactive experiences leveraging Virtual Reality, Augmented Reality and Mixed Reality have been developed to support education and training of first responders, volunteers, students, etc.

Three thesis proposals are available, to be developed mostly at Politecnico di Torino in collaboration with the said institutions.

1) The first thesis proposal aims at extending several previous thesis works that have created a prototype strategic planning and debriefing tool for forest firefighting based on a sand table. The prototype already combines a fire simulator with AR projection and mobile app-based interaction. Planned extension envisage, e.g., the integration of depth scan reconstruction of the sand construction, and the use of wearable AR headsets (HoloLens) for collaborative discussion.

2) The second thesis's gol is to build a multi-layer, multi-player training tool by integrating a series of previously-developed interactive experiences. For instance, immersive VR-based applications for training on the use of low-flame firefighting tools or high-pressure pumping equipment have been created already. The idea would be to make it possible for decision makers to see the effects of various teams operating in such applications from a kind of centralized control room.

3) The third thesis will focus on the evolution of a VR game based on a "travel-in-time" dynamics created to teach middle school students the correct procedures to follow in emergency situations. So far, hydrogeological risks (precisely floods) have been addressed. Other risks will be considered now.

In all the thesis works, experiments with end-users will be carried out to validate the effectiveness of devised solutions.

Several collaborations are in place with various institutions (Italian Airforce, Piedmont Region Civil Protection and Forest Firefighting Unit, Frejus Tunnel Authority, Fire and Rescue Department of the Savoy Region, France, etc.) involved in emergency management. In this context, a number of serious games and interactive experiences leveraging Virtual Reality, Augmented Reality and Mixed Reality have been ...

A "virtual laboratory" is defined as an interactive environment that enables a group of researchers located around the world to work together for creating and conducting simulated experiments. Virtual laboratories are receiving an increasing attention also by the world of education, as they could represent the practical side of classes that are implemented today according to the e-learning paradigm. The latter statement has become particularly relevant especially during the lockdown phase, and the potential of this technology could be finally unleashed now that person and distance education have started to be delivered at the same time in a hybrid mode.

In this context, several thesis proposals are available that are aimed to create multiple virtual laboratory experiences supporting, among others, the educational activities at Politecnico di Torino. Contacts have been activated with colleagues in various Departments interested in developing learning tools in the field of physics, chemistry, math, etc. that can be accessed by students through both Virtual Reality (VR) and Augmented Reality (AR) technologies. Various directions could be investigated, encompassing online and offline contents, self-learning modules, teacher-student and student-student collaboration modalities, etc.

An example of such tools developed, e.g., for the Department of Energy in the context of electrical engineering is presented in the paper below.

Immersive Virtual Reality for procedural training: Comparing traditional and learning by teaching approaches
Computers in Industry, January 2023
https://dx.doi.org/10.1016/j.compind.2022.103785

A "virtual laboratory" is defined as an interactive environment that enables a group of researchers located around the world to work together for creating and conducting simulated experiments. Virtual laboratories are receiving an increasing attention also by the world of education, as they could represent the practical side of classes that are implemented today according to the e-learning paradigm. The latter ...

Cyber-sickness is a disorder characterized by symptoms like nausea or discomfort, that can happen meanwhile or after using Virtual Reality (VR) technologies. Similarly to motion sickness, it is mainly caused by inconsistency between visual and aural stimuli from the simulated virtual environment and the expected feedback from the vestibular system. Several mitigation and prevention techniques have been proposed and adopted in many commercial products. However, most of them are highly situational, or require implementation choices that have a negative impact on core aspects such immersion, naturalness, and sense of presence.

In order to support a comprehensive evaluation and fair comparison of such techniques, a testbed including a set of four immersive scenarios able to stimulate different, controlled levels of cyber-sickness has been developed in a previous thesis.

The aim of this new thesis will be to study the state-of-the-art of cyber-sickness mitigation and prevention techniques by implementing them in the above testbed, and to possibly design new and more effective strategies for addressing this major issue of immersive Virtual Reality.

Cyber-sickness is a disorder characterized by symptoms like nausea or discomfort, that can happen meanwhile or after using Virtual Reality (VR) technologies. Similarly to motion sickness, it is mainly caused by inconsistency between visual and aural stimuli from the simulated virtual environment and the expected feedback from the vestibular system. Several mitigation and prevention techniques have been proposed ...

Generating computer animations, particularly of virtual characters, is a very labor-intensive task, which requires animators to operate with sophisticated interfaces. Hence, researchers continuously experiment with alternative interaction paradigms that could possibly ease the above task. Among others, the use of Virtual represents a valid alternative to traditional interfaces (based on mouse and keyboards) since it can make interactions more expressive and intuitive. However, although the literature proposes several solutions leveraging VR to solve different computer graphics challenges, generally these solutions are not fully integrated into the computer animation pipeline or were not designed to support collaborative working. Moreover, most of the existing solutions are still based on the common hand controllers, even though the literature has already shown the benefits provided by alternatives such as sketch-based interfaces [1].

The objective of this thesis proposal (more than one student will be accepted) is to design and develop an immersive and collaborative animation system that allows animators to produce and edit 3D character animations considering alternative interfaces. The system could be also featured with intelligent mechanisms able to support the animator in the creation of the virtual character animations, e.g., to automatically generate the next/previous frame given a current character pose. Related works have been proposed in the literature. For example, in [2] a system was proposed by the Disney research lab to support collaborative animations through a mixed reality system. Additional examples for authoring contents, 3D modeling, or directing actors in virtual scenes have been proposed in [3-5], however, according to a recent survey [6], it was not possible to identify works that make use of immersive and collaborative environments to produce computer animations. This research gap is probably due to the number of challenges that researchers have to face when approaching the design of collaborative AR/VR systems [7].

Related works:

[1] Cannavò, A., Zhang, C., Wang, W., Lamberti, F. (2020, October). Posing 3D Characters in Virtual Reality Through In-the-Air Sketches. In International Conference on Computer Animation and Social Agents (pp. 51-61). Springer, Cham.

[2] Pan, Ye, and Kenny Mitchell (Disney Research). "PoseMMR: a collaborative mixed reality authoring tool for character animation." 2020 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW). IEEE, 2020

[3] Nebeling, Michael, et al. "Xrdirector: A role-based collaborative immersive authoring system." Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems. 2020.

[4] Larpkiattaworn, Noppasorn, et al. "Multiuser Virtual Reality for Designing and Editing 3D Models." International Conference on Human-Computer Interaction. Springer, Cham, 2020.

[5] Wang, Cheng Yao, et al. "VideoPoseVR: authoring virtual reality character animations with online videos." Proceedings of the ACM on Human-Computer Interaction 6.ISS (2022): 448-467.

[6] Schäfer, Alexander, Gerd Reis, and Didier Stricker. "A Survey on Synchronous Augmented, Virtual and Mixed Reality Remote Collaboration Systems." ACM Computing Surveys (CSUR) (2021).

[7] Krauß, Veronika, et al. "Current practices, challenges, and design implications for collaborative AR/VR application development." Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. 2021.

Generating computer animations, particularly of virtual characters, is a very labor-intensive task, which requires animators to operate with sophisticated interfaces. Hence, researchers continuously experiment with alternative interaction paradigms that could possibly ease the above task. Among others, the use of Virtual represents a valid alternative to traditional interfaces (based on mouse and keyboards) ...

Extended reality applications are getting commonplace in several domains, ranging from entertainment to education. Moreover, the increasing interest in the Metaverse is attracting, more and more, the attention of researchers and companies. In the Metaverse, different communities can meet each other, with the aim of exchanging ideas or knowledge and working together on a given task. But what happens if someone missed an online event held in the recent past, or is interested in meeting people who lived in the previous century, etc.?

The aim of this thesis is to develop an extended reality-based system giving users the possibility to experience seamless physical events even though they happened in the past. In this way, when attendance is or was not possible, the system will be able to capture and reconstruct co-located or remote interactions, both physical and virtual, thus allowing immersed users to revisit them at a suitable time. Intelligent systems could be leveraged to make famous people get alive or reconstruct historical events, with the aim of making them available in the Metaverse.

This capability of a system to reconstruct interactions and events is generally referred to as “asynchronous reality”. The system will also allow users to transition along the reality-virtuality continuum and collaborate with other users living the same experience through different technologies. In this way, the benefits of both virtual and augmented reality can be leveraged in the same application. Developing such kind of asynchronous- and cross-reality systems is usually characterized by numerous challenges, which will be addressed in this thesis. The outcomes of this research could be applied to different use cases, such as education and training, cultural heritage, entertainment, sport, etc.

Suggested readings:

[1] Gruenefeld, Uwe, et al. "VRception: Rapid Prototyping of Cross-Reality Systems in Virtual Reality." CHI Conference on Human Factors in Computing Systems. 2022.

[2] Wang, Nanjia, and Frank Maurer. "A Design Space for Single-User Cross-Reality Applications." Proceedings of the 2022 International Conference on Advanced Visual Interfaces. 2022.

[3] Kikuchi, Yusuke, et al. "Mobile Cross Reality (XR) Space for Remote Collaboration." Human Factors in Virtual Environments and Game Design 50 (2022): 25.

[4] Woodworth, Jason W., David Broussard, and Christoph W. Borst. "Redirecting Desktop Interface Input to Animate Cross-Reality Avatars." 2022 IEEE Conference on Virtual Reality and 3D User Interfaces (VR). IEEE, 2022.

[5] Fender, Andreas Rene, and Christian Holz. "Causality-preserving asynchronous reality." Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems. 2022.

Extended reality applications are getting commonplace in several domains, ranging from entertainment to education. Moreover, the increasing interest in the Metaverse is attracting, more and more, the attention of researchers and companies. In the Metaverse, different communities can meet each other, with the aim of exchanging ideas or knowledge and working together on a given task. But what happens if someone ...

Extended Reality (XR) technologies, encompassing Virtual (VR), Augmented (AR) and Mixed (MR) Reality, are disrupting the way films are produced and viewed. Although these technologies are getting commonplace, they are presenting new, unexplored challenges to both the actors and final viewers. For instance, when motion capture (mocap) is used to record actors’ movements with the aim of animating digital characters, an increase in the workload is observed for people on stage as they have to largely rely on their imagination to understand how the digitally created characters will behave in the scene. An interesting research direction pertain how to use XR to support mocap actors’ job. Another research direction regards the impact of innovative cinematic techniques on the viewers' experience. For instance, a key step for immersive movies would be the selection of ideal points of view that would make the viewer have the best experience while watching them.

In this context, several thesis proposals are available. For instance, one of the proposals could concern the improvements of an AR system designed to support mocap actors while rehearsing scenes containing visual effects. Another proposal could focus on cinematic AR/VR, and envisage in-depth analyses aimed at supporting the creation of immersive content best matching the viewers’ expectations.

Suggested readings:

[1] Gödde M, Gabler F, Siegmund D, Braun A. Cinematic narration in VR—rethinking film conventions for 360 degrees, pp. 184–20, 2018.

[2] Cannavò A, Pratticò F G, Bruno A, Lamberti F. "AR-MoCap: Using augmented reality to support motion capture acting". In: 30th IEEE Conference on Virtual Reality and 3D User Interfaces (IEEE VR 2023), 2023.

[3] Cannavò A, Castiello A, Pratticò F G, Mazali T, Lamberti F. "Immersive Movies: The Effect of Point of View on Narrative Engagement". AI & Society, 2023.

Extended Reality (XR) technologies, encompassing Virtual (VR), Augmented (AR) and Mixed (MR) Reality, are disrupting the way films are produced and viewed. Although these technologies are getting commonplace, they are presenting new, unexplored challenges to both the actors and final viewers. For instance, when motion capture (mocap) is used to record actors’ movements with the aim of animating digital ...

Several thesis proposals are available regarding the use of haptic feedback in Virtual Reality, Augmented Reality and Mixed Reality experiences. The proposals are in collaboration with different teams of the IRISA-Inria center at Rennes university and include an internship at their premises (http://www.irisa.fr/en/scientific-departments-irisa). Duration of the thesis/internship is approximately 5-6 months. Experience in C/C++/C#, Unity3D, VR/AR tools, human-machine interaction are required. Details are provided below at at the following link:
inria-irisa-rennes-haptics-thesis-internship-april2023.pdf


HapticMic: Studying the Persuasive Effects of Haptics during Speech-Based Interactions in Virtual Reality

Vibrotactile feedback is directly related to sound. When we speak, we make our body vibrate. At a concert, we feel vibratory feedback if we get close to a speaker. Vibrations seem to be an interesting way to emphasize sound feedback. Persuasive speech is broadly used in verbal communication, for instance in meetings, advertisements, informal discussions among friends, etc. Being able to modulate speaker leadership or persuasion in a collaborative environment could be of interest to increase inclusivity (e.g. of shy participants) or to solve conflicts. Previous studies have suggested that leadership is influenced by visual feedback. Very recently, we conducted two experiments where participants embody a first-person avatar attending a virtual meeting in immersive VR. Results showed that vibrotactile-reinforced speech can significantly improve the perceived co-presence but also the persuasiveness and leadership of the haptically-augmented agent. The objective of this research work is to design and evaluate a haptic-enabled microphone, able to reinforce speech in real-time through vibrotactile feedback.


Wearable haptics for augmenting tangible objects in Augmented Reality

Augmented Reality (AR) integrates virtual content into our real-world surroundings, giving the illusion of one unique environment. Virtual object manipulation is critical for useful and effective AR usage, such as in medical applications, training, or entertainment. However, there still exist several issues that affect manipulation in AR, degrading the overall user experience. One of the most important limitation is the lack of haptic feedback, despite being proven to significantly improve both the manipulation performance and the user's experience during manipulation in VR and AR. The lack of haptic sensations is due to multiple reasons, including the degradation of the tracking performance when using, e.g., wearable haptic interfaces, and the resulting mismatch between the pose of the user's hand and that of its avatar representation in the AR environment. The objective of this research work is to design and evaluate (wearable) haptic approaches to augment tangible objects during AR manipulation, e.g., alter the haptic perception of a tangible object through additional wearable haptic stimuli.


Mixed Reality (MR) for heterogeneous multi-robot systems

Teams of coordinated robots have been successfully used in a plethora of different applications, including disaster response, exploration, patrolling, and surveillance. Multi-robot systems are able to perform actions according to the perception of the single robot and understanding of the environment, as the team (physically) interacts with it. The ensemble of these abilities of sensing, interpreting, modeling, predicting, and interacting with the physical world are concrete applications for Artificial Intelligence (AI) tools and methodologies. However, when multiple robots are involved, it is unclear how to provide spatially accurate and efficient information to a (leader) human user, who either control the team from a remote (and secure) location or moves in the same environment as the robots. In this respect, Mixed and Augmented reality can be used to provide the human user with additional information about the target environment and task, reconstructing (part of) it from the information provided by the robotic team. By enhancing the world with virtual information, the user is allowed to provide additional feedback to the robots, that can elaborate those information using sensors and cameras in a complete agnostic way with respect to the environment they move in. The objective of this work is to design and deploy a mixed-reality system able to agnostically treat real and virtual robots, as well as provide feedback information to the user about the status of the robots. Imagine a human user, moving in the same environment of a team of coordinated drones; he or she can see directly some of the drones, while those who are not visible are shown through AR rendering. Moreover, virtual content can be added to the real world to show additional relevant details, such as the location of targets.


Immersive Virtual Reality (VR) multi-sensory User Interface (UI) for controlling multi-robot systems at the microscale

Information feedback during the teleoperation of microrobotic systems is of course fundamental, enabling the human user to understand what is happening in the remote environment. An example of visual feedback commonly provided during the teleoperation of such systems is shown in Figure 1: two microrobots are controlled in 3D, providing the user with the front and side views of the environment as captured by the cameras. Input information regarding, e.g., where the robot should move is usually provided by simply clicking on the screen to indicate the target reference positions. While this approach is widely employed, and it provides all the necessary information to drive the robots, it does not enable an intuitive control and visualization of the robots and their environment. The objective of this research work is to design and evaluate a series of immersive Virtual Reality (VR) User Interfaces (UI) for controlling multi-robot systems at the microscale in an intuitive and natural way, as well as providing feedback information to the user about the status of the robots and the current task.

Several thesis proposals are available regarding the use of haptic feedback in Virtual Reality, Augmented Reality and Mixed Reality experiences. The proposals are in collaboration with different teams of the IRISA-Inria center at Rennes university and include an internship at their premises (http://www.irisa.fr/en/scientific-departments-irisa). Duration of the thesis/internship is approximately 5-6 months. ...

Reconstruction of the complete kinematics of a collision between particles at accelerators is a fundamental step in the investigation of subtle processes hidden in the 10-15 m and 10-23 s space-time windows where the interaction takes place. The study of pion – Helium nucleus interaction at the temperature at which a transition phase from a bound nucleus to a gas of fermions is expected to occur is of high interest because it allows to investigate a number of fundamental open issues in nuclear and particle physics. It can provide information on the Universe evolution during the transition phase from the hadron era to the primordial nucleosynthesis era (10-3 s, 1011 K - 107 s, 109 K). During this transition the strong force (Quantum ChromoDynamics) has been responsible for a non-trivial set of physics phenomena involving, at the same time, the emergence of collective self-interacting complex systems existing at the quantum scale.

At the Joint Institute for Nuclear Research (JINR) (Dubna-Russia) a set of several thousand pion - Helium nucleus scattering events has been observed with a detector, namely a self-shunted streamer chamber (SSSC) filled with Helium, capable of revealing the complete interaction kinematics, down to the momenta of very low-energy charged secondary interaction products such as protons with an energy of merely 1 MeV. A charged particle leaves along its trajectory electron-ion pairs, that give rise to electric discharges in their initial stage (streamers), produced when a high-voltage pulse is applied to the SSSC electrodes, if a signal from scintillation counters reveals a nuclear event to have occurred inside the chamber volume. The visible light, emitted in excitation and ionization processes taking place within the streamers in the Helium gas filling the volume of the detector, indicates the particle tracks and is captured by two CCDs and saved in two photograms. The kinematics (the momenta of particles that take part in the collision) is reconstructed from the analysis of the tracks left by charged particles.

This thesis will focus on the 3D Reconstruction and VR visualization of nuclear scattering events produced and photographed in the PAINUC experiment at the JINR phasotron (particle accelerator) of the Laboratory of nuclear problems (LNP). In particular, the work will be focused on the implementation of automatic 3D reconstruction of each pion - Helium nucleus scattering event, on the basis of the two photograms taken by the CCDs. The work will involve the use of sw tools for the stereoscopic view reconstruction as well as for background reduction and filtering. The effectiveness of the various tools, from standard image filtering approaches to the use of AI-based systems, will be studied.

In addition, visualization of the reconstructed events in a VR environment is to be implemented in order to provide a new tool for the analysis of phenomena involved as well as a powerful environment focused on STEM didactics and the Communication of Science.

The thesis will be developed in collaboration with CERN, JINR and DISAT.

Reconstruction of the complete kinematics of a collision between particles at accelerators is a fundamental step in the investigation of subtle processes hidden in the 10-15 m and 10-23 s space-time windows where the interaction takes place. The study of pion – Helium nucleus interaction at the temperature at which a transition phase from a bound nucleus to a gas of fermions is expected to occur is of high ...

The aim of this thesis, to be developed in collaboration with the Egyptian Museum in Turin, is to develop an algorithm able to reconstruct in a digital way the fragments of archeological remains by analyzing the fracture surfaces of the single pieces. The algorithm could be first tested on a “modern” object reduced to pieces, and then applied to a use case to be agreed with the Museum (e.g., stone fragments from a statuary group, decorated floor or ceiling pieces, ceramic parts, etc.). The match between two fragments shall be identified based on the analysis of the geometries, not of the possible color information that could still be visible on the surface. The algorithm should provide a confidence score for the various matches identified. 3D geometries will be provided by the Museum. These geometries will be created using the Artec Space Spider, a structured light scanner capable to produce meshes with a resolution up to 0.1 mm.

Suggested readings:
https://www.sciencedirect.com/science/article/abs/pii/S1296207422000826
https://ieeexplore.ieee.org/document/4697243
https://www.researchgate.net/publication/333161346_Reassembling_Fractured_Objects_by_Geometric_Matching
https://www.academia.edu/51368373/Automated_reconstruction_of_fragmented_objects_using_their_3D_representation_Application_to_important_archaeological_finds
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3231500/

The aim of this thesis, to be developed in collaboration with the Egyptian Museum in Turin, is to develop an algorithm able to reconstruct in a digital way the fragments of archeological remains by analyzing the fracture surfaces of the single pieces. The algorithm could be first tested on a “modern” object reduced to pieces, and then applied to a use case to be agreed with the Museum (e.g., stone fragments ...

One of the challenges for future intelligent vehicle applications, will be to determine whether a person is authorized to operate the vehicle or not. The face is the most common identifier in humans. Face recognition is a biometric application to recognize and identify the facial patterns of the human faces. These facial patterns of persons are compared and extracted by their facial contours. Anyway, extraneous imaging components, for example, posture, light and outward appearance still reason much inconvenience in exact face acknowledgment.

Furthermore, if the problem of recognizing the owner of a vehicle moves outside the car itself, that is, the system must recognize the owner in exceptional access conditions, the problem becomes even more tough - recognition in the “wild”. It means to build a fast and accurate system that can detect, recognize, and verify the driver’s identity in presence of constraints introduced in the car environment in different lighting and other external conditions. Recently, human pose estimation is an important and widely concerned research topic in computer vision. Given an image or video input, 3D human pose estimation aims to predict the configuration of the human body. The use of body posture can contribute to biometric identification with a multimodal approach.

The focus of this thesis, developed in collaboration with Centro Ricerche Fiat (CRF) and Stellantis, will be on multimodal AI models, combining face identification with body pose estimation to tackle the above challenges.

One of the challenges for future intelligent vehicle applications, will be to determine whether a person is authorized to operate the vehicle or not. The face is the most common identifier in humans. Face recognition is a biometric application to recognize and identify the facial patterns of the human faces. These facial patterns of persons are compared and extracted by their facial contours. Anyway, extraneous ...

This thesis falls within the activities of the AIBIBANK (Bio-Banking for Artificial Intelligence) project, carried out in collaboration with HealthTriage srl (I3P). The overarching goal is to develop artificial intelligence solution for the automated triage of digital mammography images in order to focus the radiologists’ attention to potentially normal cases. Since cancer is a rare disease, the dataset available is highly skewed towards normal cases. In addition, the content of mammography images depend on the vendor (manufacturer) of the mammographic unit, and the characteristics of the woman (age, breast density, etc.). In order to make the models more robust to under-represented categories, the aim of this thesis is the development of generative models capable of synthesizing high resolution images. One or more of the following objectives will be pursued: - developing models able to convert (with techniques such as CycleGAN) the global characteristics of the image, such as changing the vendor, the breast side or density
- generate and inject synthetic lesions in normal cases, from the same or different vendors, using inpanting techniques. Multiple views of the same lesion should be generated.
- experiment with the use of models based on Stable Diffusion for the generation of high resolution mammography images Strong programming and analytical skills are required. Experience with deep learning and Pytorch is required.

Suggested readings
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6430964/pdf/JMI-006-031411.pdf
https://www.frontiersin.org/articles/10.3389/fonc.2022.868257/full
https://arxiv.org/pdf/2209.09809.pdf

This thesis falls within the activities of the AIBIBANK (Bio-Banking for Artificial Intelligence) project, carried out in collaboration with HealthTriage srl (I3P). The overarching goal is to develop artificial intelligence solution for the automated triage of digital mammography images in order to focus the radiologists’ attention to potentially normal cases. Since cancer is a rare disease, the dataset ...

The main goal of this thesis is to design and implement a proof of concept for extracting information items (such as date, gross amount, and tax code) from scanned semi-structured documents (invoices) in close collaboration with a leading insurance company. The proposed system will exploit recent advances in self-supervised vision-language-layout transformer models that can analyze each element based on their content (graphical and textual) as well as their position and relationship with respect to the rest of a document [1,2]. Starting from a generic pre-trained model on the English language, the system will be fine-tuned to the task of Information Extraction on Italian documents. Performance will be evaluated based on standard and ad-hoc metrics for information extraction reflecting (a) the percentage of documents for which information can be sufficiently extracted and (b) the accuracy of the extracted information. Experience with deep learning and Pytorch is a prerequisite. Good programming and analytical skills are required.

Suggested reading:
[1] Unifying Vision, Text, and Layout for Universal Document Processing. https://arxiv.org/pdf/2212.02623.pdf
[2] OCR-free document understanding transformer. https://arxiv.org/pdf/2111.15664.pdf

The main goal of this thesis is to design and implement a proof of concept for extracting information items (such as date, gross amount, and tax code) from scanned semi-structured documents (invoices) in close collaboration with a leading insurance company. The proposed system will exploit recent advances in self-supervised vision-language-layout transformer models that can analyze each element based on their ...

The ultimate goal of this activity is to explore the potential use of large language models (LLMs) and chatbots, such as GPT-2 and GPT-3, in the insurance domain. Multiple thesis are available to investigate one or more of the following topics: a) how to fine-tune an existing LLM, starting from open source and commercial solutions, to customize it for a target language (Italian) and domain (insurance/legal); b) exploiting the LLM to analyze insurance claims. Example of target tasks are: predict whether they should be reimbursed under a given policy, or classify the type of damages described. Different alternatives will be considered, including prompt engineering and traditional classifiers ; c) exploit how to use structured Knowledge Graphs (KGs) in order to fine-tune and/or expand the capabilities of a LLM. Experience with deep learning and Pytorch is a prerequisite. Good analytical skills are required. Prior NLP knowledge is preferred, but not required.

The ultimate goal of this activity is to explore the potential use of large language models (LLMs) and chatbots, such as GPT-2 and GPT-3, in the insurance domain. Multiple thesis are available to investigate one or more of the following topics: a) how to fine-tune an existing LLM, starting from open source and commercial solutions, to customize it for a target language (Italian) and domain (insurance/legal); b) ...

Current social media generate a tremendous amount of visual material, that can be exploited by researchers operating in social media research, digital humanities, and marketing. However, privacy regulations impose significant restrictions to both data collection and sharing. The CAMOUFLAGE project, funded by the AI4Media European project, aims to exploit diffusion model (such as Stable Diffusion) to produce a synthetic version of an existing image, presenting equivalent visual and semantic characteristics of the original, while at the same time fully preserving the anonymity of the user who published the image. The project has a strong multidisciplinary focus and will be carried out in collaboration with researchers in visual semiotics to determine the quality and information content of the generated image. The system will be tested on a collection of real Facebook and Instagram profile pictures. Multiple thesis are available tackling three distinct, yet related, research objectives: to design and implement controllable image synthesis that retains the visual and semantic content of a target image; to determine whether the resulting synthetic images can be considered successfully anonymized, in comparison with state-of-the-art solutions based on GANs; and whether the synthetic collection has the same information content of the original image.

Current social media generate a tremendous amount of visual material, that can be exploited by researchers operating in social media research, digital humanities, and marketing. However, privacy regulations impose significant restrictions to both data collection and sharing. The CAMOUFLAGE project, funded by the AI4Media European project, aims to exploit diffusion model (such as Stable Diffusion) to produce a ...

Two thesis proposals are available in the field of Artificial Intelligence and Digital Humans, to be developed in collaboration with Reply Spa. The aim of the first thesis work is to create synthetic avatars to be included in the metaverse and in business contexts; the plan is to use Natural Language Processing (NLP) via tools like DialogFlow and GPT3 to make the conversation on a given topic/subject natural, realistic and interactive.

References:
DialogFlow
https://cloud.google.com/dialogflow/ cx/docs/basics
OpenAI
https://beta.openai.com/docs/
Digital humans
https://digitalhumans.com/


The goal of the second thesis work is to develop agents that can assist customers in completing Cooking and DIY tasks by incorporating multimodal (voice and screen) conversational interface with AI; to this aim, it will be necessary to combine screen-based interaction into the conversational experience so that, in addition to receiving verbal instructions, screen devices could also be used with step-by-step instructions, images, or diagrams that enhance task guidance.

References:
Alexa Prize TaskBot Challenge:
https://www.amazon.science/alexa- prize/taskbot-challenge/rules

Two thesis proposals are available in the field of Artificial Intelligence and Digital Humans, to be developed in collaboration with Reply Spa. The aim of the first thesis work is to create synthetic avatars to be included in the metaverse and in business contexts; the plan is to use Natural Language Processing (NLP) via tools like DialogFlow and GPT3 to make the conversation on a given topic/subject natural, ...

Two thesis proposals in collaboration with Reply Spa are available in the context of self-supervised learning (SSL) and computer vision applied to mobile robots. In particular, the aim of the first thesis work is to develop new methods of SSL in computer vision in order to minimize the bottlenecks and problems of the classical methods of supervised and unsupervised learning. The goal is to adopt and improve an SSL model and then insert it into a real use case in industry or customer.

References:
Masked Siamese Networks for Label-Efficient Learning
https://arxiv.org/abs/2204.07141
SSL method DINO
https://github.com/facebookresearch /dino


The objective of the second thesis work is to apply computer vision models to robotic agents such as Spot, by Boston Dynamics, or parcel delivery robots and other IOT devices. The difficulty of this work lies in being able to compress the power of the SSL models into lighter models capable of giving excellent results having scarce reserves both hardware and connectivity.

References:
Tensorflow lite
https://www.tensorflow.org/lite
Robot Operating System
https://www.ros.org/

Two thesis proposals in collaboration with Reply Spa are available in the context of self-supervised learning (SSL) and computer vision applied to mobile robots. In particular, the aim of the first thesis work is to develop new methods of SSL in computer vision in order to minimize the bottlenecks and problems of the classical methods of supervised and unsupervised learning. The goal is to adopt and improve ...

Artificial Intelligence agents are required to learn from their surroundings and to reason about the knowledge that has been learned in order to make decisions. While state-of-the-art deep learning typically uses distributed representations, reasoning is normally useful at a higher level of abstraction. As a result, attempts at combining symbolic AI and neural computation into neural-symbolic systems have been on the increase. Neural-symbolic learning and reasoning is an area of research that aims at integrating relational/symbolic reasoning and knowledge representation (e.g., first order logic language for knowledge representation) by embedding their constructs in deep neural networks. Image interpretation tasks, such as object detection, image classification or visual relationship detection, may benefit from neural-symbolic integration in order to represent and exploit high-level information about objects and their relationship. The goal of this project is to design deep architectures combining convolutional neural networks with neural-symbolic components such as Logic Tensor Networks to solve various image interpretation tasks. One of the potential advantages of this emerging technique is the possibility to exploit prior information available in the form of semantic networks, e.g. WordNet or ConceptNet, to improve the learning process. Multiple thesis proposals are available tackling the following problems: i) techniques for end-to-end training of deep neuro-symbolic networks, ii) integration of prior knowledge in the training process, and iii) characterization of neural-symbolic architectures, e.g. in terms of types of features learnt, efficiency and interpretability. Skills possessed or to be acquired: programming skills (Python, Keras/Tensorflow, or other deep learning framework); good analytical and mathematical skills.

Suggested readings:
https://arxiv.org/pdf/1705.08968
https://arxiv.org/pdf/2012.13635

Artificial Intelligence agents are required to learn from their surroundings and to reason about the knowledge that has been learned in order to make decisions. While state-of-the-art deep learning typically uses distributed representations, reasoning is normally useful at a higher level of abstraction. As a result, attempts at combining symbolic AI and neural computation into neural-symbolic systems have been ...

Multiple thesis projects are available within the context of an EU-funded project dedicated to investigating the meaning of the face in present-day digital cultures following a cross-disciplinary approach and in collaboration with a multidisciplinary team. The project focuses on how users represent their identity in the digital world and in particular on social networks. Other sources of images that may be considered are images of faces in art and history, and artificial facial images generated by GANs. Within this project, activities to be carried out span from the collection of images (in compliance with GDPR) to their analysis using a variety of tools from computer vision, deep learning and visual analytics. Image analysis will be performed through the extraction of quantitative parameters and ad-hoc trained deep convolutional neural networks. Examples of quantitative parameters extracted from the facial images are presence of faces (profiles pictures may represent other types of objects), facial pose, facial landmarks, etc. An analytics platform, to be developed, will be used to characterize the distribution of image properties in space and time. The activities are expected to start in March 2021. The exact activities within each thesis will be defined based on the starting period and interactions with other members of the research team. Strong programming skills are expected. Previous knowledge of deep learning, computer vision or big data analytics are required.

Multiple thesis projects are available within the context of an EU-funded project dedicated to investigating the meaning of the face in present-day digital cultures following a cross-disciplinary approach and in collaboration with a multidisciplinary team. The project focuses on how users represent their identity in the digital world and in particular on social networks. Other sources of images that may be ...

Many End-to-End (E2E) testing tools allow developers to create repeatable test scripts that mimic a human user's interaction with the application and evaluate its response. Various paradigms of testing tools have been defined that are differentiated based upon the way the widgets are located on the GUI (e.g., text, images, layout properties). This thesis falls within the visual GUI testing paradigm scope, where screen snapshots used as visual locators:
- identify the widgets, and
- replicate an existing test suite across multiple devices or versions of the same app.

This technique leverages image recognition algorithms.

In particular, the thesis focuses on testing mobile applications where the same widgets have to be identified across multiple devices, with different screen sizes and characteristics. In this context, marginal variations in the graphical rendering can invalidate widgets' recognition if the image recognition algorithm is not sufficiently robust. This thesis aims to extend an existing prototype for visual GUI testing by leveraging deep learning. In particular, the following activities are envisioned:
- Training a detector to identify widgets and their classes with high recall.
- Using the features extracted by the neural network to match the same widgets when the app is rendered across multiple devices.
- Investigating the possibility of integrating text extraction with image analysis.

The solution will be characterized and possibly optimized in terms of accuracy and execution speed. The developed techniques will be included in an open-source library.
The student should possess or should be willing to acquire these skills:
- programming skills (Python, deep learning frameworks);
- experience in training deep neural networks;
- fundamentals of mobile development (Android GUI, the Android Studio development environment).

Many End-to-End (E2E) testing tools allow developers to create repeatable test scripts that mimic a human user's interaction with the application and evaluate its response. Various paradigms of testing tools have been defined that are differentiated based upon the way the widgets are located on the GUI (e.g., text, images, layout properties). This thesis falls within the visual GUI testing paradigm scope, where ...

Training deep neural networks require large scale annotated datasets which are often difficult to collect in the medical domain. As a consequence, a plethora of small and very specialized datasets have emerged. The medical domain would extremely benefit from more general purpose models that can be applied to a variety of problems. A possible strategy to overcome this issue is to leverage self-supervised techniques in which the network is pre-trained by using a pretext task. A simple example is rotation: the image is rotated and the network learns to detect the orientation; more generally, a pretext task can be any for which the labels are automatically generated. If the pre-training dataset is sufficiently large and varied, the features learnt should be sufficiently general to be useful for many downstream tasks. The network is then fine-tuned towards the final task (e.g., lesion detection or segmentation). The goal of this thesis, starting from an existing dataset generated in a previous work, is to systematically explore different self-supervised learning techniques, including generative components, and explore their relationship with the diagnostic/classification tasks. Secondly, we will explore ways to make the features learnt generalize better to different imaging modalities. Strong programming and analytical skills are required, as well as a solid background in machine learning/deep learning.

Suggested readings:
https://arxiv.org/abs/1908.06912
https://arxiv.org/abs/1912.01991

Training deep neural networks require large scale annotated datasets which are often difficult to collect in the medical domain. As a consequence, a plethora of small and very specialized datasets have emerged. The medical domain would extremely benefit from more general purpose models that can be applied to a variety of problems. A possible strategy to overcome this issue is to leverage self-supervised ...

Multiple thesis are available within the AIBIBANK (Bio-Banking for Artificial Intelligence) project. Research activities at DAUIN, developed in collaboration with HealthTriage srl, aim at developing deep learning-based components for the automated screening of digital mammography images. In early cancer prevention programs, women undergo biennal exams in which up to four, high-resolution images are acquired. The goal of the envisioned system is to triage women to identify cases with high probability of being normal, for expedite reading, and secondarily identify lesions. The problem is technically and scientifically challenging since mammography images are much larger than traditional RGB images, and subtle information from multiple views must be integrated in a single prediction; the trained networks must operate in a robust and interpretable ways. Within the project, different deep learning architectures (based on convolutional networks, transformers and relational networks) will be compared. Each thesis will focus on one or more of the following problems: i) design and comparison of multi-stream architectures can process multiple images at once, at different resolutions/levels of detail, ii) methods to estimate the uncertainty of trained neural networks, iii) data augmentation, data harmonization, self-supervised pre-training and domain generalization techniques to facilitate robust training across different vendors, and iv) strategies to enhance the interpretability of the trained networks and to incorporate clinical knowledge during training or in the post-hoc explanation phase. Students will be involved in a multi-disciplinary research team. Strong programming and analytical skills are required, as well as a solid background in machine learning/deep learning. Previous experience with medical imaging is a plus, but not required.

Suggested readings:
Evaluation of Combined Artificial Intelligence and Radiologist Assessment to Interpret Screening Mammograms
https://pubmed.ncbi.nlm.nih.gov/32119094/

High-Resolution Breast Cancer Screening with Multi-View Deep Convolutional Neural Networks
http://arxiv.org/abs/1703.07047

Multiple thesis are available within the AIBIBANK (Bio-Banking for Artificial Intelligence) project. Research activities at DAUIN, developed in collaboration with HealthTriage srl, aim at developing deep learning-based components for the automated screening of digital mammography images. In early cancer prevention programs, women undergo biennal exams in which up to four, high-resolution images are acquired. ...

Batch normalization has been a key enabler in the training of very deep neural networks. However, it suffers from a number of drawbacks including slow updates, instability with small batch sizes, different behaviors at training and inference times, and platform-dependent implementations than hinder research reproducibility. Recently, variations on batch normalization techniques, including batch-norm-free architectures, were proposed. However, most recent papers focus experiments on high-end, multi-GPU training environments, that enable very large batch sizes; these platforms are beyond the reach of many practitioners. The goal of this thesis is to investigate and compare different training strategies in single-GPU environments, with small to medium batch sizes, and to determine the best architectural and training strategies for low-resource training environments. Strong programming and analytical skills are required. Previous experience with at least one deep learning framework (Tensorflow/Keras or Pytorch) is required, as the thesis may involve developing new (or modifying existing) layers and optimizers.

Suggested reading:
High-Performance Large-Scale Image Recognition Without Normalization
https://arxiv.org/pdf/2102.06171.pdf

Batch normalization has been a key enabler in the training of very deep neural networks. However, it suffers from a number of drawbacks including slow updates, instability with small batch sizes, different behaviors at training and inference times, and platform-dependent implementations than hinder research reproducibility. Recently, variations on batch normalization techniques, including batch-norm-free ...

Deep neural networks, such as convolutional neural networks, are sensitive to the presence of dataset bias, which arises when the distribution of training samples is skewed and conflates the presence of unrelated attributes. For instance, photos of wolves may be associated more frequently with a snowy background, photos of cats are typically taken indoor, photos of bikes are typically associated with sunny weather, certain professions are typically associated to a certain gender, and so forth. As a consequence, the network may learn to “take shortcuts” and rely on undesired attributes to make predictions, and fail to generalize on unseen instances. For instance, CNNs may make predictions based on the background, rather than the object, or may base their predictions on undesirable attributes (such as gender). Typically, this is corrected by verifying the predictions (e.g., through post-hoc explanations), adjusting the dataset (e.g., by acquiring new data) and re-training the network. The goal of this thesis is to investigate neural-symbolic techniques through which deep neural networks can be trained to “comply” with desiderata in the form of logical constraints (e.g., “a doctor can be a men or women with equal probability”) to prevent dataset bias from affecting the training process. Strong programming and analytical skills are required. Experience with Tensorflow/Keras is preferred or will be acquired.

Suggested reading:
Faster-LTN: a neuro-symbolic, end-to-end object detection architecture
https://arxiv.org/abs/2107.01877
Shortcut learning in deep neural networks
https://arxiv.org/abs/2004.07780

Deep neural networks, such as convolutional neural networks, are sensitive to the presence of dataset bias, which arises when the distribution of training samples is skewed and conflates the presence of unrelated attributes. For instance, photos of wolves may be associated more frequently with a snowy background, photos of cats are typically taken indoor, photos of bikes are typically associated with sunny ...

The driving style is unique to each person and it affects the vehicle usage. The thesis aims to create an autonomous agent able to mimic the behavior of a specific driver, in order to perform adaptive DSE (Drive Style Estimation). This requires the agent to be able both to react properly to environmental stimuli and to tailor its behavior on the driver’s profile. Activities could encompass the use of the both datasets collected on real vehicles, as well as data gathered in simulated environments.

The thesis will be carried out in collaboration with Sensor Reply.

The student will be involved in: - Vehicle data analysis and data preprocessing;
- Development of an AI-powered autonomous agent;
- Adapt the agent to the specific driver.


Relevant technologies: Deep Learning, Adversarial Learning, Reinforcement Learning, Autonomous Agent, Edge AI, Cloud Computing, Connected Vehicle

The driving style is unique to each person and it affects the vehicle usage. The thesis aims to create an autonomous agent able to mimic the behavior of a specific driver, in order to perform adaptive DSE (Drive Style Estimation). This requires the agent to be able both to react properly to environmental stimuli and to tailor its behavior on the driver’s profile. Activities could encompass the use of the both ...

Reconstruction of the complete kinematics of a collision between particles at accelerators is a fundamental step in the investigation of subtle processes hidden in the 10-15 m and 10-23 s space-time windows where the interaction takes place. The study of pion – Helium nucleus interaction at the temperature at which a transition phase from a bound nucleus to a gas of fermions is expected to occur is of high interest because it allows to investigate a number of fundamental open issues in nuclear and particle physics. It can provide information on the Universe evolution during the transition phase from the hadron era to the primordial nucleosynthesis era (10-3 s, 1011 K - 107 s, 109 K). During this transition the strong force (Quantum ChromoDynamics) has been responsible for a non-trivial set of physics phenomena involving, at the same time, the emergence of collective self-interacting complex systems existing at the quantum scale.

At the Joint Institute for Nuclear Research (JINR) (Dubna-Russia) a set of several thousand pion - Helium nucleus scattering events has been observed with a detector, namely a self-shunted streamer chamber (SSSC) filled with Helium, capable of revealing the complete interaction kinematics, down to the momenta of very low-energy charged secondary interaction products such as protons with an energy of merely 1 MeV. A charged particle leaves along its trajectory electron-ion pairs, that give rise to electric discharges in their initial stage (streamers), produced when a high-voltage pulse is applied to the SSSC electrodes, if a signal from scintillation counters reveals a nuclear event to have occurred inside the chamber volume. The visible light, emitted in excitation and ionization processes taking place within the streamers in the Helium gas filling the volume of the detector, indicates the particle tracks and is captured by two CCDs and saved in two photograms. The kinematics (the momenta of particles that take part in the collision) is reconstructed from the analysis of the tracks left by charged particles.

This thesis will focus on the automatic recognition of particle tracks in nuclear scattering events taken by the PAINUC experiment at the JINR phasotron (particle accelerator) of the Laboratory of nuclear problems (LNP).

A purpose of this thesis consists in the determination of distributions of various kinematic parameters (such as particle momenta, energies, effective masses etc.) of intermediate-energy pion – Helium nucleus events for possible revelation of subtle processes between nucleons in the Helium nucleus. To this end the thesis work will be focused on automatic 3D reconstruction of the trajectories (tracks) of initial and secondary particles pertaining to nuclear interaction events occurring in the LNP self-shunted streamer chamber (SSSC) filled with Helium. Over 30 thousand such events have been collected during runs of the SSSC in pion beams of the LNP phasotron. Pion – Helium nucleus events were photographed using CCDs that registered visible light emitted by excited helium atoms in streamers along particle tracks; the light is gathered in voxels (3D pixels). Identification of the tracks will be followed by the determination of a set of fundamental parameters characterizing the interactions: the vertex (where the pion – Helium nucleus scattering occurred) and the best fitting functions for describing the identified tracks (typically a 3D helix of variable radius). The system will also provide for the identification of non-interacting particles and other noise sources. The task will be achieved by applying both standard pattern recognition algorithms (i.e. Cellular Automata) and AI-based approaches. The efficiency of the automatic identification algorithm will be essential for the subsequent analysis of the physics phenomena hidden in the scattering event kinematics as well as for providing a powerful didactic tool for introducing nuclear and particle physics to students and the general public.

The thesis will be developed in collaboration with CERN, JINR and DISAT.

Reconstruction of the complete kinematics of a collision between particles at accelerators is a fundamental step in the investigation of subtle processes hidden in the 10-15 m and 10-23 s space-time windows where the interaction takes place. The study of pion – Helium nucleus interaction at the temperature at which a transition phase from a bound nucleus to a gas of fermions is expected to occur is of high ...

The thesis aims at designing and developing a tool for the analysis of smart contracts deployed on existing blockchain platforms. The objective of the tool is to support the user in the identification of smart contracts' main characteristics (e.g., in terms of application field, type, tasks performed, interactions with other smart contracts, etc.), by also exploiting visualization techniques. The thesis student will also investigate the added value of search engines for smart contracts, and will enrich the developed tool with an "intelligent" search feature, e.g., based on semantic technologies.

The thesis aims at designing and developing a tool for the analysis of smart contracts deployed on existing blockchain platforms. The objective of the tool is to support the user in the identification of smart contracts' main characteristics (e.g., in terms of application field, type, tasks performed, interactions with other smart contracts, etc.), by also exploiting visualization techniques. The thesis student ...

The adoption of electric vehicles (EVs) is increasing, and the demand for electric vehicle recharging infrastructure is growing. To ensure the efficient operation of this infrastructure, new tools are needed to simulate, test, and optimize different charging strategies. In this thesis proposal, we propose the design and implementation of a blockchain-based simulation tool for electric vehicle recharging infrastructure that can:
- provide an immutable and decentralized record of all charging transactions and other relevant data;
- exploit different consensus mechanism to model, simulate, and test different charging strategies for electric vehicle recharging infrastructure.

In the first part of the thesis the student will be asked to develop a blockchain oracle represented by the porting in Python of an existing MATLAB model.

In the second part, the student will design and develop the blockchain platform based on an Ethereum-like (e.g., Quorum, Polygon) or an Hyperledger (e.g., Fabric, Besu) blockchain.

In the final phase the oracle will be connected to the blockchain platform to test and validate the integrated simulation tool.

The methodology for this thesis will include both theoretical and practical research. The theoretical research will involve a thorough analysis of existing blockchain consensus and their limitations, as well as a study of dynamic models and external oracles. The practical research will involve the design, implementation and testing of the simulation tool composed by the external oracle and the blockchain platform, as well as the evaluation of its performance and security. The thesis will be developed in collaboration with LINKS Foundation.

The adoption of electric vehicles (EVs) is increasing, and the demand for electric vehicle recharging infrastructure is growing. To ensure the efficient operation of this infrastructure, new tools are needed to simulate, test, and optimize different charging strategies. In this thesis proposal, we propose the design and implementation of a blockchain-based simulation tool for electric vehicle recharging ...

Self-Sovereign Identity is a new digital identity model that empowers individuals to control their identity without relying on centralized authorities. SSI enables the creation, management, and verification of digital identities that are self-owned, portable, and interoperable. In the context of an energy data marketplace, SSI can potentially enable secure and efficient sharing of energy data while giving individuals control over their data.

This thesis aims to explore the concept of Self-Sovereign Identity (SSI) and its potential integration within an energy data marketplace. The purpose of the research is to analyze the algorithms used to enable the methodology of SSI, and investigate the implications of adopting SSI in an energy data marketplace. This research aims to contribute to the body of knowledge on the use of SSI in various applications and provide a practical demonstration of its implementation in an energy data marketplace.

In particular, the research will begin with a comprehensive review of the literature on SSI and its algorithms and will examine its applicability to an energy data marketplace. The study will investigate the technical and social implications of SSI integration in the marketplace and identify potential benefits and limitations.

The research will then proceed to identify the most suitable SSI algorithms for the energy data marketplace and investigate the feasibility of implementing them.

The thesis will be developed in collaboration with LINKS Foundation.

Self-Sovereign Identity is a new digital identity model that empowers individuals to control their identity without relying on centralized authorities. SSI enables the creation, management, and verification of digital identities that are self-owned, portable, and interoperable. In the context of an energy data marketplace, SSI can potentially enable secure and efficient sharing of energy data while giving ...

Industrial product tracking is a critical aspect of supply chain management that ensures the integrity, traceability, and security of products as they move through the supply chain. Blockchain technology has the potential to revolutionize industrial product tracking by providing a tamper-proof and decentralized system for data storage and traceability. However, with various blockchain platforms available, it is important to evaluate the suitability of each platform for industrial product tracking based on both technical and business aspects.

This thesis aims to conduct a benchmarking of blockchain platforms for the traceability of a specific industrial prototyping process using a SWAT (Strengths, Weaknesses, Opportunities, Threats) analysis considering relevant technical and business aspects (e.g., as cost alignment, sustainability, scalability, longevity, integration, regulatory compliance, and access to blockchain development skill) and demonstrating the advantages of a blockchain-based solution for industrial production traceability through the development of a proof-of-concept.

The specific objectives if the thesis work include:
- Research and analyze existing blockchain platforms for industrial product tracking
- Conduct a SWAT analysis of the blockchain platforms considering relevant technical and business aspects
- Evaluate the suitability of each blockchain platform for industrial product tracking based on the results of the SWAT analysis
- Provide recommendations for the most suitable blockchain platform for industrial product tracking based on the findings.
- Implement a proof-of-concept using one of the selected platform to demonstrate the effectiveness of a blockchain-based approach to for industrial production traceability.

The expected results of this thesis include a comprehensive benchmarking of blockchain platforms that will provide insight into the suitability of each blockchain platform for industrial product tracking and recommendations for the most suitable blockchain platform based on the findings, together with an actual implementation of a blockchain-based solution to demonstrate the feasibility of the proposed solution.

The thesis will be developed in collaboration with LINKS Foundation.

Industrial product tracking is a critical aspect of supply chain management that ensures the integrity, traceability, and security of products as they move through the supply chain. Blockchain technology has the potential to revolutionize industrial product tracking by providing a tamper-proof and decentralized system for data storage and traceability. However, with various blockchain platforms available, it is ...