PhD Program in  Computer Science and Systems Engineering 

Research Projects Proposals 2019 (XXXV Cycle)

Research line: Data Science and Engineering

Title: Joint segmentation, detection, tracking in video sequences for efficient and effective scene understanding

Description: Intrinsically of a different nature, image segmentation and object detection address two different questions with the common goal of understanding the content of a scene. Image segmentation is often seen as a lower level tasks. In recent years we assisted to the development of semantic segmentation approaches, where we also associate semantic labels with pixels, super-pixels, or image regions. In this way we obtain an overall understanding of the scene, as image regions may contain objects as well as background areas. Semantic segmentation methods are providing today good performances at the price of being computationally demanding,  for this reason  in video analysis they are usually applied at a lower rate and propagated on the following frames. If the camera is moving propagation requires motion estimation. Conversely detection is an efficient, higher level task, that applies to entities associated with a defined spatial extent (objects);  object detection is usually less accurate in terms of localisation (it provides bounding boxes). For its higher level nature it is easier  to extend it to video analysis through the use of state of the art tracking or prediction algorithms.

We will explore different ways of combining these complementary sources of information, trying to achieve a good compromise between efficiency and effectiveness. Our focus will be primarily on the analysis of video sequences acquired by moving cameras, there included applications related with robotics and automation and autonomous guidance.

Title: Machine learning for prognostic maintenance
Proposers: Stefano Rovetta, Francesco Masulli
Curriculum: Computer Science

Description: Predictive maintenance is widely acknowledged as the “killer application” of machine learning in Industry 4.0.  This research activity will develop machine learning methods for prognostic maintenance, an approach that aims not only at predicting the future maintenance necessities, but also at describing causes and effects of future evolutions of a system: “foresight,” as opposed to “forecast.”

The  activity will be carried on in collaboration with a software company that already markets a more traditional solution for predictive maintenance. Therefore, the work will build on an existing, substantial body of tools and know-how. The candidate is expected to develop competences that are of great technical, industrial, ans well as commercial, interest.

Link to the group or personal webpage: https://www.dibris.unige.it/rovetta-stefano

References:

[1] Vogl, G.W., Weiss, B.A. & Helu, M. “A review of diagnostic and prognostic capabilities and best practices for manufacturing.” J Intell Manuf (2019) 30: 79.

Title: Smart request processing for personalized data space-user interactions through approximation and learning

Proposers: Barbara Catania, Giovanna Guerrini
Curriculum: Computer Science

Description:
The increase of data size and complexity requires a deep revisiting of user-data interactions and a reconsideration of the notion of query itself. A huge number of applications need user-data interactions that emphasize user context and interactivity with the goal of facilitating interpretation, retrieval, and assimilation of information [1]. The ability to learn from observations and interactions [2] as well as to approximate process requests [3,4] are two key ingredients in these new settings.

The aim of this research theme is to devise smart innovative approaches for exploiting, from a processing viewpoint and a focus on graph-shaped data, the role of user context (geo-location, interests, needs) and of similar requests repeated over time to inform approximation and refine knowledge on underlying data, which in turn can be used to more efficiently and effectively fulfill information needs. Preliminary approaches in this direction have been proposed in [5].

Link to the group or personal webpage: dama.dibris.unige.it

References:

[1] Georgia Koutrika. Databases & People: Time to Move on From Baby Talk. EDBT/ICDT ‘18

[2] Yongjoo Park, Ahmad Shahab Tajik, Michael Cafarella, and Barzan Mozafari. Database Learning: Toward a Database that Becomes Smarter Every Time. ACM SIGMOD ‘17

[3] Peter Haas. Some Challenges in Approximate Query Processing. EDBT/ICDT ‘2018

[4] Barbara Catania, Giovanna Guerrini. Adaptively Approximate Techniques in Distributed Architectures. SOFSEM ‘15

[5] Barbara Catania, Francesco De Fino, Giovanna Guerrini. Recurring Retrieval Needs in Diverse and Dynamic Dataspaces: Issues and Reference Framework. EDBT/ICDT Workshops 2017

Title: Making motion analysis computationally efficient

Proposers: Nicoletta Noceti, Giuseppe Ciaccio
Curriculum: Computer Science

Short Description: Motion analysis is one of the main elements in Computer Vision, at the basis of a variety of higher-level tasks, as activity recognition and behavior understanding. The earliest task of motion analysis is always the ability to identify the moving regions in image streams, often referred to as motion-based image (and video) segmentation.  Two main scenarios can be identified:

• When the videos are acquired with a still camera, the task can be formulated as a pixel-based classification (moving or still?)
• When there is no prior information on the setting of acquisition, the problem of motion-based image segmentation is intertwined with motion estimation.

In these fields, most of the work in the recent decades has a focus on accuracy of results; only a marginal share of the literature deals with performance, intended as execution time of the algorithms, and efficiency, intended as power consumption of the computing device running a specific algorithm. The two latter aspects, however, play a critical role in real-world visual applications, due to increasing frame resolution/throughput requirements and increasing demand for stand-alone battery-powered autonomous device applications (e.g. robots).

We will therefore investigate the possibility of leveraging modern highly-parallel computer architectures (e.g. many-core processors) and throughput-oriented devices (e.g. GPUs) along with power-constrained platforms (e.g. FPGAs or low-power computers), using state-of-the-art numerical methods and code optimization techniques, in order to cope with the growing computational demands of motion analysis.  The goal of this project is to devise methods for motion-based image segmentation, with particular focus on background subtraction and optical flow estimation, specifically intended to be effective but also performing and efficient.

References

• Sundaram, N., Brox, T., & Keutzer, K. (2010). Dense point trajectories by GPU-accelerated large displacement optical flow. In European conference on computer vision(pp. 438-451). Springer, Berlin, Heidelberg.
• Stagliano, A., Noceti, N., Verri, A., & Odone, F. (2015). Online space-variant background modeling with sparse coding. IEEE Transactions on Image Processing, 24(8), 2415-2428.

Title: Analysis of Genic Expression via Game Theory and Graphs

Proposer: Marcello Sanguineti
Curriculum: Computer Science 

Short description. Recent research (e.g., Albino et al. (2008)) have emphasized the possibility of applying Game Theory (Peters (2008) to the analysis of medical results obtained via the so-called “microarray techniques”. Such techniques allow one to “take a picture” the expression of thousands of genes in a cell by means of a single experiments. The departure point is the study of the genetic expression in a sample of cells that satisfy particular biological conditions: for instance, cells belonging to a cancer-affected subject. Game Theory plays a basic role in defining the “microarray games” (Schena et al. (1995)) and in evaluating the relevance of each gene in influencing or even determining a pathology, by taking into account the interactions with other genes. To this end, in the literature (see, e.g., Moretti et al. (2007)) it has been investigated the use of some “power indices” from Game Theory (such as the Shapley and Banzhaf values) to estimate gene relevance. In particular, an in-depth analysis has been performed in relationship with colon cancer and neuroblastic tumors.

The aim of this research project consists in studying the use of Game Theory in the analysis of genic expression data and contributing to a better understanding of some originating factors of cancer. As a first step, the already-available studies based on the Shapley and Banzhaf values in microarray games will be further developed and tested on case-studies made available the research group of Prof. Alberto Ballestrero at IRCCS at S. Martino Hospital in Genova, with whom we have already established a joint research plan. The partner team from S. Martino has access to databases made up of some hundreds of microarray experiments and RNA-sequencing, made available by the Cancer Genome Atlas Network (TCGA). As a second step, the research aims at considering other power indices, such as the “τ-value”. The third phase will be devoted to a comparison among the various indices, both from theoretical and experimental viewpoints, in such a way to evaluate which are the better-suited in the study of cancer. Finally, the combination of tools from Game Theory with graph-based approaches, which allow one to model interactions between pairs of genes, will be examined.

Link to the group/personal webpage:

http://www.dist.unige.it/msanguineti/

References

-Albino D., Scaruffi P., Moretti S., Coco S., Di Cristofano C., Cavazzana A., et al. . Identification of low intratumoral gene expression heterogeneity in neuroblastic tumors by wide-genome analysis and game theory. Cancer 113(6):1412-1422, 2008.

-Moretti S., Patrone F., Bonassi S. The class of microarray games and the relevance index for genes. Top 15:256-280, 2007.

-Peters H. Game Theory. A Multi-Leveled Approach, Springer, 2008.

-Schena M., Shalon D., Davis W.R., Brown P.O. Quantitative Monitoring of Gene Expression Patterns with a Complementary DNA Microarray. Science 270:467-470, 1995.

Research line: Artificial Intelligence and Multiagent Systems

Title: Hybrid deliberative/reactive robot architectures for task scheduling and joint task/motion planning 

Proposers: Marco Maratea (This email address is being protected from spambots. You need JavaScript enabled to view it.), Fulvio Mastrogiovanni (This email address is being protected from spambots. You need JavaScript enabled to view it.)
Curriculum: Computer Science

Short description: A longstanding goal of Artificial Intelligence (AI) is to design robots able to autonomously and reliably move in the environment and manipulate objects to achieve their goals. Autonomy and reliability can be considered still unsolved issues when robots operate in everyday environments, especially in the presence of humans. Traditionally, beside specific activities in perception, knowledge representation, action, and the mechanical structure of robots, an important research trend is related to the architecture robots may adopt to enforce autonomy and reliability, and specifically robustness and resilience to unexpected events, as well as uncertainty in perception and action outcomes.

This project aims at investigating, designing and prototyping robot architectures able to (i) interleave scheduling (i.e., the long-term definition of what a robot should do in the future) and task planning (i.e., which specific actions a robot should perform next), and (ii) integrate task and motion planning (i.e., what robot trajectories correspond to planned actions), in full perception-representation-reasoning-action loop.

On the one hand, the integration between scheduling and planning has not received sufficient attention in the literature, and only recently the issue has been studied, possibly relying on modeling through logical languages, e.g., PDDL or Answer Set Programming.

On the other hand, while discrete task planning has been considered mainly in the AI community, continuous motion planning has been the focus in much Robotics research. Such a separation leads to suboptimal robot behavior in real-world scenarios, especially in case of not modeled events, misperceptions or uncertainty in sensory data. However, in the past few years, a number of approaches have been discussed in the literature, which aim at integrating the discrete and the continuous planning process. The recent introduction of such planning formalisms as PDDL+ is a decisive step in this direction, and its effective use in Robotics architectures has not been fully explored yet.

The Ph.D. student will be involved in ongoing research activities in the application of advanced AI techniques to Robotics. In particular, the following topics will be considered:

• Definition of expressive and computationally efficient knowledge representation approaches for robots.
• Definition of innovative and efficient scheduling algorithms suitable for a robotic setting.
• Representation of robot perceptions using logic formalisms able to ground further reasoning processes, i.e., knowledge revision, update, fusion.
• Design and implementation of joint task/action planning strategies for robots.

Link to the group/personal page: http://www.star.dist.unige.it/~marco/, https://www.dibris.unige.it/mastrogiovanni-fulvio

Title:Inductive and Deductive Reasoning in Transportation

Proposers: Davide Anguita (This email address is being protected from spambots. You need JavaScript enabled to view it.), Marco Maratea (This email address is being protected from spambots. You need JavaScript enabled to view it.)

Macro-areas: Artificial Intelligence, Data Analysis

Curriculum:Computer Science

Short description:Inference is defined as the act or process of deriving logical conclusions from premises known or assumed to be true. Deduction is an inference approach since it does not implies any risk. Once a rule is assumed to be true and the case is available there is no source of uncertainty in deriving, through deduction, the result. Inductive reasoning, instead, implies a certain level of uncertainty since we are inferring something that is not necessary true but probable. The inductive reasoning is the simple inference procedure that allows to increment our level of knowledge since the induction allows to infer something that is not possible to logically deduce just based on the premises. New generation of information systems collects and store a large amount of heterogeneous data which allow to induce Data Driven Models able to forecast their evolution. Deep, Multi-Task, Transfer, Semi-Supervised learning algorithms together with rigorous statistical inference procedures allow us to transform large and heterogeneous amounts of distributed and hard-to-interpret pieces of information in meaningful, easy to interpret and actionable information. Data Driven Models scale well with the amount of data available but they are not as effective if exploited for deduction purposes. On the contrary, Model Based Reasoning allows to model in an effective way complex systems based on the physical knowledge about them and deduce meaningful information by solving complex (optimization) problems. The general idea is to encode an application problem as a logical specification. This specification is subsequently evaluated by a general-purpose solver, whose answers can be mapped back to solutions of the initial application problem. The Model Based Reasoning limitation is that they may not scale well with the size of problem. The scope of this PhD proposal is to make inductive and deductive reasoning work together for the purpose of solving real world problems related to the transportation domain (e.g. Railway, Busses, and Airways). Transport of goods and people is a multifaceted problem since it involves technical constraints coming from the limited physical assets, safety constraints, social and cultural implication. In Europe the increasing volume of people and freight transported is congesting the transportation systems. The challenge of this research theme is twofold. From one side there is a necessity to exploit and further refine the state-of-the-art tools and basic research themes in the inductive and deductive fields in order to make induction and deduction work together and solve the respective limitations. From the other side there are plenty of real world problems in public transportation (e.g. multimodal transportation systems, train dispatching, combinatorial problems, and forecast problems) that needs the combination of different technological tools and techniques in order to be able to obtain satisfying results.

Link to the group/personal page: www.smartlab.ws,http://www.star.dist.unige.it/~marco/

Title: Neural Software Engineering

Proposers: Massimo Narizzano,  Armando Tacchella
Curriculum: Computer Science

Description:  Neural networks are one of the most investigated and widely used techniques in Machine Learning. In spite of their success, they still find limited application in safety- and security-related contexts, wherein assurance about networks’ performances must be provided. In the recent past, automated reasoning techniques have been proposed by several researchers to close the gap between neural networks and applications requiring formal guarantees about their behavior. This research program starts with a comprehensive categorization of existing approaches for the automated verification of neural networks. The purpose of this phase is to understand current limitations and directions for investigation in this topic at the crossroads of Machine Learning and Automated Reasoning. The second phase seeks to provide an evaluation of current approaches from a quantitative point of view: what can be verified, what are the time/space requirements, what are the fundamental limitations that must be overcome. Finally, the program seeks to design and implement a tool which is capable to analyse neural networks in a purely automated fashion, in order to uncover adversarial examples and provide hints to fix network weights to eliminate them.

Link to the group or personal webpage: www.aimslab.org

References:

Luca Pulina, Armando Tacchella: An Abstraction-Refinement Approach to Verification of Artificial Neural Networks. CAV 2010: 243-257
Luca Pulina, Armando Tacchella: NeVer: a tool for artificial neural networks verification. Ann. Math. Artif. Intell. 62(3-4): 403-425 (2011)
Luca Pulina, Armando Tacchella: Challenging SMT solvers to verify neural networks. AI Commun. 25(2): 117-135 (2012)

Title: Configuration and Optimization of Complex Systems

Proposer: Armando Tacchella
Curriculum: Computer Science

Description:  Computer-automated design (CautoD) differs from “classical” computer-aided design (CAD) in that it is oriented to replace some of the designer’s capabilities and not just to support a traditional work-flow with computer graphics and storage capabilities. While CautoD programs may integrate CAD functionalities, their purpose goes far beyond the replacement of traditional drawing instruments and most often involves the use of advanced techniques from artificial intelligence. As mentioned in [BOP+16], the first scientific report of CautoD techniques is the paper by Kamentsky and Liu [KL63], who created a computer program for designing character-recognition logic circuits satisfying given hardware constraints. In mechanical design — see, e.g., [RS12] — the term usually refers to tools and techniques that mitigate the effort in exploring alternative solutions for structural implements, and this is the flavor of CautoD that will be considered hereafter.  The goal of this research program is to evaluate general purpose AI methods to solve and optimize configurations of complex systems. Starting from the work presented in [AMT17] related to elevators, the idea is to consider configuration software (e.g., CLIPS) and other constraint-based techiques (e.g., Constraint Programming) in order to configure systems (or parts thereof) automatically starting from specifications.

Link to the group or personal webpage: http://www.aimslab.org/

References:

[AMT17] Leopoldo Annunziata, Marco Menapace, Armando Tacchella: Computer Intenstive vs. Heuristic Methods in Automated Design of Elevator Systems. In 31st European Conference of Modelling and Simuation, ECMS 2017 (to appear).

[BOP+16]Robin T. Bye, Ottar L. Osen, Birger Skogeng Pedersen, Ibrahim A. Hameed, and Hans Georg Schaathun. A software framework for intelligent computer-automated product design. In 30th European Conference on Modelling and Simulation, ECMS 2016, Regensburg, Germany, May 31 – June 3, 2016, Proceedings., pages 534–543, 2016.

[KL63] Louis A. Kamentsky and Chao-Ning Liu. Computerautomated design of multifont print recognition logic. IBM Journal of Research and Development, 7(1):2–13, 1963.

[RS12] R. Venkata Rao and Vimal J. Savsani. Mechanical design optimization using advanced optimization techniques. Springer Science & Business Media, 2012.

Research line: Secure and Reliable Systems

Research line: Human-Computer Interaction 

Title: Study and Development of Innovative Methods and Technologies for Analysis, Recognition, Quantification, Co-Registration, Fusion of Ultrasound and/or MRI Images to Support Diagnostics

(Funded by CNR and ESAOTE SPA, according to the CNR-Confindustria Agreement 2018)

Proposers: Dr. Giuseppe Patane’ (IMATI-CNR,http://www.imati.cnr.it), Dr Andrej Dvorak (ESAOTE SPA https://www.esaote.com)
Curriculum: Computer Science

Research environment: The training, research, and development activities will involve CNR-IMATI, as Research Partner, and ESAOTE SPA as Industrial Partner, with a collaboration with main Hospitals in Genova and with the Ligurian Regional Hub on Life Sciences.

Short Description The aim of this Industrial Ph.D. is the training of researchers and multi-skilled professionals capable of covering roles that require a deep knowledge of methodological and operational aspects in Computer Science, Computer Graphics, and Bio-medicine, with a focus on the modeling, visualization, and analysis of biomedical images acquired by ultrasound and/or magnetic resonance techniques. The training of the Ph.D. student will provide a solid scientific and technological background on Computer Science, Information Engineering, and Bio-medical Research. Particular attention will be paid to the main aspects of research and development for an Industrial Ph.D., in terms of scientific publications, development, and patents. The proposed research project falls within the thematic area of Computer Science, Biomedical Sciences, and advanced technological research in the area of information technology for health and biomedical industry, with a particular focus on the early diagnosis, screening, therapy, and follow-up of pathologies. Furthermore, the project has a multidisciplinary value, with fundamental and applied research in the fields of 3D Computer Graphics, Bio-medicine, Scientific Visualization, and Medicine.

References

• Mattia Natali, Giulio Tagliafico, Giuseppe Patanè: Local up-sampling and morphological analysis of low-resolution magnetic resonance images. Neurocomputing 265: 42-56 (2017).
• Imon Banerjee, Giuseppe Patanè, Michela Spagnuolo: Combination of visual and symbolic knowledge: A survey in anatomy. Comp. in Bio. and Med. 80: 148-157 (2017).

Title: Perception-based rendering for virtual reality environments

Proposer: Fabio Solari
Curriculum: Computer Science

Short Description: Virtual reality provides complex and realistic worlds, where users can interact to perform several tasks.  A fundamental aspect of interactions is how humans perceive virtual worlds, thus visual perception studies can play an important role: modeled perception can suggest how to modify the rendering of the virtual world in order to adaptively optimize it for human observers and for their perceptual bandwidths. For instance, stereo cues (e.g. disparity) are predominant in peripersonal interactions, motion (e.g. optic flow) and perspective cues are essential in navigation and walking tasks, and space variant human sight effectively modulates spatial resolution. These perceptual cues could be used to devise and to implement visualization techniques that exploit them to render virtual environments (and also virtual representations of real objects that users can touch) in a natural way by allowing users to improve their performance and mitigate fatigue.  Thus, the research theme aims in particular to develop perception-based advanced rendering techniques, which can be used to improve the user’s experience in virtual worlds.

Link to the group/personal webpage: https://www.dibris.unige.it/en/solari-fabio

https://sites.google.com/site/fabiosolarifs/

Title: Multimodal interactive systems based on non-physical dimensions of touch

Proposers: Antonio Camurri, Enrico Puppo, Davide Anguita, Gualtiero Volpe

Description This PhD proposal aims at investigating computational models and developing techniques and systems for the automated measure of tactility: how non-verbal, social, affective content usually communicated and perceived by touch can be communicated and perceived without any physical contact. Can tactility be as effective as the physical one in socio-emotional communication? Scientific research (e.g., McKenzie et al., 2010) as well as artistic theories and practice (e.g., the dance) demonstrate the existence of tactility. Humans are able to perceive touch even in cases of lack of physical contact, since movement alone may induce in an observer the perception of touch. Touch conveys emotions, facilitates or enhances compliance in social interactions. Touch reduces the negative effects of several chronic disease. Illusory touch occurs when people believe they have been touched but no actual tactile stimulation has been applied. This PhD proposal focuses on computational models of tactility, that is, to study and develop systems to enable the communication and perception of non-verbal, social, affective content usually communicated and perceived by touch, but without any physical contact. Tactility is the carrier of non-verbal emotional and social communication. Research challenges include the following: how does an observer perceive tactility and its role in socio-emotional interaction? Does an observer of tactility performed on a “ghost” body perceive the same socio-affective message as on a physical body?

Proposed work plan
– Assessment of the interdisciplinary existing state of the art: motion capture, biomechanics, crossmodal perception (Spence 2011), humanistic theories and computational models of non-verbal multimodal full-body movement analysis (Kleinsmith & Berthouze 2016), social signal processing (Vinciarelli et al 2012), analysis of 3D trajectories, machine learning. Software environments for the development of real-time multimodal systems (EyesWeb http://www.infomus.org/eyesweb_ita.php);
– Design of a dataset and of pilot experiments. The dataset consist of a pool of movements performed by a number of pairs or small groups of participants highly skilled in movement execution (e.g., dancers) as well as poorly skilled. For example, two participants in front of each other at a few steps of distance; the first slowly walks to approach and touch the other (e.g. on a shoulder); then she returns to the original position, the second leaves the scene, and the first repeats the same action and touches the “ghost” of the second participant: she touches the memory, a sort of tactile photography.

The dataset will be recorded using the Qualysis (www.qualisys.com) motion capture and other sensor systems (physiology, IMUs, audio) available at DIBRIS premises of Casa Paganini-InfoMus;
– Analysis of tactility: extraction of a collection of multimodal features from the recorded data that explain the difference of same touch gesture on a real human Vs the “ghost”;
– Assessment of the analysis outcomes by comparison with ratings of the same dataset provided by human participants;
– Development, evaluation and validation of prototypes of multimodal systems exploiting tactility.

Expected results
– A collection of an archive of MoCap and multimodal data for the analysis of tactility, to be made publicly available to the research community;
– Development of novel algorithms, techniques, and software libraries for the automated analysis of tactility;
– Scientific publications in top-level international conferences and journals;
– Development of prototypes of systems exploiting tactility in at least one of the following scenarios: therapy and rehabilitation in specific activities of the ARIEL (Augmented Rehabilitation) Joint Laboratory DIBRIS-Gaslini Children Hospital; enhanced active experience of cultural heritage in collaboration with Palazzo Reale Museum in Genoa;
– Participation to public dissemination events: e.g., European Commission events, international workshops and conferences, summer schools, science festivals;
– The research may be part of international projects, including European funded Horizon 2020 ICT projects, running at Casa Paganini-InfoMus research centre.

Link to the group or personal Webpage:

www.casapaganini.org www.youtube.com/InfoMusLab dance.dibris.unige.it

Casa Paganini – InfoMus Research Centre publications:  http://www.infomus.org/publications_ita.php

References
– Camurri, A., & Volpe, G. (2016). The Intersection of art and technology. IEEE MultiMedia, 23(1), 10-17.
– Kleinsmith, A., Bianchi-Berthouze, N. (2013). Affective body expression perception and recognition: A survey. IEEE Transactions on Affective Computing, 4(1), 15-33.
– McKenzie, K. J., Poliakoff, E., Brown, R. J., and Lloyd, D. M. (2010). Now you feel it, now you don’t: how robust is the phenomenon of illusory tactile experience? Perception, 39(6), 839-850.
– Spence, C. (2011). Crossmodal correspondences: A tutorial review. Attention, Perception, & Psychophysics, 73(4), 971-995.
– Vinciarelli, A., Pantic, M., Heylen, D., Pelachaud, C., Poggi, I., D’Errico, F., & Schroeder, M. (2012). Bridging the gap between social animal and unsocial machine: A survey of social signal processing. IEEE Transactions on Affective Computing, 3(1), 69-87.

Title: Techniques for the design and implementation of (Spatial) Augmented Reality Environments

Proposer: Manuela Chessa
Curriculum : Computer Science

Short Description Augmented Reality (AR) allows a real-time blending of digital information (e.g. text, virtual elements, images, sounds) onto the real world. Among the different technologies to design and implement AR scenarios, we can distinguish between wearable devices (e.g. the HoloLens) and non-wearable solutions, in particular Spatial Augmented Reality (SAR). This approach allows displaying additional information, virtual objects, or even changing the appearance of the physical objects directly onto the real environment. It is worth noting that,  compared to head mounted displays AR displays or handheld devices, SAR has some advantages, e.g. it allows the interaction with physical (yet augmented) objects and it scales well to multiple users and therefore supports collaborative tasks naturally. On the other side, many issues are still open, e.g., a robust detection of the 3D structure of the environments and the registration between virtual and real contents. Moreover, the combination with handled or wearable devices can further improve the range of possible solutions and interaction systems.

The research theme aims to develop novel techniques to create AR and SAR environments in which people can interact in an ecological way. Besides virtual visual information added to the real world, also sensorized objects providing controlled force and tactile feedbacks could be used to augment reality and devise novel interaction paradigms.

Link to the group/personal webpage: 

www.dibris.unige.it/en/chessa-manuela

www.manuelachessa.it

Research line: Systems Engineering

Title:  Sustainable planning and control of distributed power and energy systems

Proposers: R. Minciardi, M. Robba
Curriculum: Systems Engineering

Description: The increase in the use of renewable energies, the emergence of distributed generation and storage systems, and, in general, the concept of “smart grids”, have given rise to the necessity of defining new decision and control schemes for planning and management purposes. Currently, a major challenge is represented by the lack of a unified mathematical framework including robust tools for modeling, simulation, control and optimization of time critical operations in complex multicomponent and multiscaled networks, characterized by microgrids, interconnected buildings, renewables, storage systems and electric vehicles. The difficulty of defining effective real time optimal control schemes derives from the structure of a power grid, and, specifically, from the presence of several issues: renewable and traditional power production, bidirectional power flows, dynamic storage systems, demand response requirements, and stochastic aspects (such as uncertainties in renewable, prices, and demand forecasting). This results in optimization problems, which are generally intractable within a real time optimal control scheme, if all components of the whole system are represented at a full level of detail. Moreover, the new regulation related to new market entrants and schemes requires a revision and improvement of distributed energy management systems planning and management, as well as their coordination in order to optimize self-consumption and energy distribution.

The proposed PhD research activity will fall within this framework and has the objective of developing and applying tractable approaches for planning and optimal control, taking into account stochastic issues (i.e., intermittent renewables, demands, prices) and considering different possible architectures (multilevel, decentralized, distributed). In particular, the formulation of the optimization and control problems will be based on realistic models for the electrical grid and for its various sub-systems (microgrids, intelligent buildings, storage systems, renewables). Moreover, different energy distribution systems will be taken into account in relation to polygenerative systems: district heating, buildings heating and cooling, with the associated storage systems, water, and gas distribution networks. Finally, different kinds of demands will be taken into account (heat, cool, electricity), as well as electrical vehicles with charging/discharging cycles within a smart grid.

References

F. Delfino, R. Minciardi,  F. Pampararo, M. Robba. A Multilevel Approach for the Optimal Control of Distributed Energy Resources and Storage, IEEE Transactions on Smart Grid, Special Issue on Control Theory and Technology in Smart Grid, to appear

S. Bracco, F. Delfino, F. Pampararo, M. Robba, M. Rossi. A mathematical model for the optimal operation of the University of Genoa Smart Polygeneration Microgrid: Evaluation of technical, economic and environmental performance indicators, Energy, 2013

H. Dagdougui, R. Minciardi, A. Ouammi, M. Robba, R. Sacile. A dynamic decision model for the real time control of hybrid renewable energy production systems, IEEE Systems Journal, Vol. 4, No. 3, 2010, p. 323-333.

Title: Optimal routing and charging of electrical vehicles in a smart grid.
Proposers: Riccardo Minciardi, Massimo Paolucci, Michela Robba
Curriculum:  System Engineering

Description: At international level, different new policies have been developed to reduce CO2 emissions, such as the Kyoto Protocol, the European 20-20-20 strategy, and the Energy roadmap 2050. The result is an increase of green technologies for energy production and transportation. Due to the presence of intermittent and distributed production (such as renewables (RES)) and loads (such as electrical vehicles (EVs)), the actual electrical grid management has to be changed, and new control strategies are necessary for the integration of electrical and transportation networks. In fact, on one side, EVs need to be charged in the fastest time possible, and, on the other side, smart grids should afford such a request. In particular, from users’ perspective it is necessary to know electrical consumes over a specific path, and to decide where and how much to charge EVs to satisfy their own travel exigencies. Instead, from the grid perspective, it is necessary to offer adequate charging facilities based on control strategies that are able to satisfy users but guaranteeing electrical grid constraints. The proposed PhD research activity will fall within this framework. In particular, the following main objectives/activities can be listed:

• Definition and development of a discrete event optimization model for microgrids with EVs.
• Definition and development of power management strategies for charging stations.
• Optimal routing and charging of EVs: development of meta- and math-heuristics.

Demonstration activities with real charging stations and case studies (in collaboration with companies) are foreseen  during the three years.

Link to personal homepage

http://www.dibris.unige.it/minciardi-riccardo

http://www.dibris.unige.it/robba-michela

http://www.dibris.unige.it/paolucci-massimo

References:

Schneider, M., Stenger, A., Hof, J., 2014, An Adaptive VNS Algorithm for Vehicle Routing Problems with Intermediate Stops, in Technical Report LPIS-01/2014.

Yagcitekin, B., Uzunoglu, M., 2016, A double-layer smart charging strategy of electric vehicles taking routing and charge scheduling into account,

Title:  Smart scheduling approaches for manufacturing industry.

Proposer: Massimo Paolucci
Curriculum:  System Engineering

Short description: Scheduling in manufacturing industry involves key decisions about how to exploit at best the available resources (e.g., machines, tool, workers, energy) in order to efficiently perform the required production activities. Scheduling decisions are at the operational level, that is, they regard a short time planning horizon (a day or shift) and must take into account detailed production conditions and requirements. In real manufacturing industries scheduling problems are at a large scale (the number of activities to be performed may be huge and workshops may include many machines and tools) so that the number of possible alternative decisions usually grows exponentially. In addition, even if scheduling problems share common features, several relevant differences exist which characterize different industrial sector (e.g., food and beverage, fashion, automotive). Therefore an effective general purpose solution approach that could represent the basis for developing scheduling systems for different sectors, avoiding to restart from scratch with a specific algorithm, seems not available. The introduction of the Industry 4.0 paradigm will allow to rely on fresh data from the field, so improving the possibility of planning, adapting and revising the scheduling decisions more effectively, even reacting to the unpredicted changes that usually characterize the real production systems. Finally, sustainability issues, as energy consumption and carbon footprint, need to be included among the scheduling objectives.

The purpose of this research project is to design a new solution approach for facing a large class of the scheduling problems emerging in manufacturing industry. Such an approach can be based on several building block and strategies (recent metaheuristics as adaptive large neighborhood search or bio-inspired algorithms, simulation-optimization as well as heuristics based on mathematical programming) that can be exploited to design a solver framework for this class of hard combinatorial problems. 

Link to the group or personal webpage:

http://www.dibris.unige.it/paolucci-massimo

Title: Strategic and tactical planning in production and logistics for manufacturing industry

Proposer(s): Massimo Paolucci
Curriculum:  System Engineering

Short description: Planning at strategic and tactical level are connected problems influencing both the design of the supply chain and the manufacturing production activities. Such decisions usually involve the activation of facilities, the allocation of the available resources, as well as the aggregate management of both production and distribution activities over a medium-long time horizon. Apparently strategic and tactical planning decisions impact not only business objectives but also environmental sustainability; as an example, in closed loop supply chains planning decisions also include the use of recovered materials and components from returned products and in general the reduction of energy consumption.

Therefore the proposed research project aims at considering the problem of planning in the supply chain for manufacturing production systems in order to define a general purpose decision support system able to operate both at strategic and tactical level. The purpose is to determine a unified model and a set of optimization approaches to support planning decisions at different levels (e.g., supply network design, inventory and lot-size planning, distribution planning), including sustainability aspects, such as remanufacturing and energy consumption and emissions. Since at least part of the considered optimization problems are computationally intractable as they belong to the NP-hard complexity class, the algorithms that need to be designed and tested can range from exact approaches, based on mathematical programming models, to heuristic, metaheuristics (from neighborhood search techniques to population and bio-inspired algorithm) or matheuristics (i.e., methods that include mathematical programming models in a heuristic solution framework).

Link to the group or personal webpage:

http://www.dibris.unige.it/paolucci-massimo

Title:  Hyper- and meta- heuristics for multi-objective optimization

Proposer: Massimo Paolucci
Curriculum:  System Engineering

Short description: Most of the decision problems in real life applications require to take into account more than one objective/criterion. Usually such objectives are non-commensurable and conflicting. These problems arise in many different fields and often express the conflict between customer satisfaction, stakeholders profit and social and environmental sustainability. As an example, in manufacturing industry and logistics, planning the activities on the supply chain should aim at timely meeting the customer demand, reducing production, inventory and transportation costs, minimizing the energy consumption and CO₂ emission, favoring material recycling and so on. Multi-criteria decision making deals with this wide class of decision problems and embeds Multi-objective optimization as those methods whose purpose is to define the set of solutions which deserve to be considered by decision makers. Such solutions are the so-called efficient or Pareto optimal ones. In general the problem of determining the Pareto optimal solutions for a multi-objective optimization problem is NP-hard and the dimension of such set of solutions is exponential. For this reason, metaheuristic algorithms, such as Genetic Algorithms, Simulated Annealing, Ant Colony Optimization and Particle Swarm Optimization, have been applied to multi-objective optimization since 90s. The purpose of this research is to deep investigate the possible use of metaheuristics for multi-objective optimization, trying in particular to design general purpose self-adapting algorithms. This can be pursued by experimenting the so-called hyper-heuristics, consisting in combining higher level metaheuristics with lower level metaheuristics, where the purpose of the former is to identify the best configuration of the latter when solving a given optimization problem.

Transportation Network Optimization Via Transferable-Utility Games

Proposer: Marcello Sanguineti
Curriculum: Systems Engineering

Short description. Network connectivity is an important aspect of any transportation network, as the role of the network is to provide the society with the ability to easily travel from point to point using various modes. Analyzing networks’ connectivity can assist the decision makers with the identification of weak components, to detect and prevent failures, and to improve the connectivity in terms of reduced travel time, reduced costs, increase reliability, easy access, etc..

A basic question in network analysis is: how “important” is each node? An important node might, e.g., highly contribute to short connections between many pairs of nodes, handle a large amount of the traffic, generate relevant information, represent a bridge between two areas, etc. To quantify the relative importance of nodes, one possible approach consists in using the concept of “centrality” [1, Chapter 10]. A limitation of classical centrality measures is the fact that they evaluate nodes based on their individual contributions to the functioning of the network. For instance, the importance of a stop in a transportation network can be computed as the difference between the full network capacity and the capacity when the stop is closed. However, such an approach is inadequate when, for instance, multiple stops can be closed simultaneously. As a consequence, one needs to refine the existing centrality measures, in such a way to take into account that the network nodes do not act merely as individual entities, but as members of groups of nodes. To this end, one can exploit game theory [2], which, in general terms, provides a basis to develop a systematic study of the relationship between rules, actions, choices, and outcomes in situations that can be either competitive or non-competitive.

The idea at the roots of game-theoretic centrality measures [3] is the following: the nodes are considered as players in a cooperative game, where the value of each coalition of nodes is determined by certain graph-theoretic properties. The key advantage of this approach is that nodes are ranked not only according to their individual roles in the network, but also taking into account how they contribute to the roles of all possible groups of nodes. This is important in various applications in which a group’s performance cannot be simply described as the sum of the individual performances of the group members involved. In the case of transportation networks, suppose we have at our disposal a certain budget. One possible approach consists in addressing the question of whether investing all the money in increasing the capacity and/or service of a transportation component (road section, bridge, transit route, bus stop, etc.)  substantially improves the whole network. A better way of proceeding for the network analyst/designer would probably consist in considering to simultaneously improve a (possibly small) subset of the components. In this case, to evaluate the importance of a component one has to take into account the potential gain of improving one component as a part of a group of components, not merely the potential gain of improving the component alone. This approach can be formalized in terms of cooperative game theory [2], where the nodes are players whose performances are studied in coalitions, i.e., subsets of players.

This research project, which takes the hint from the works [4,5], consists in developing methods and tools from a particular type of cooperative games, called “cooperative games with transferable utility”, for brevity “TU games”, to optimize transportation networks. Given a transportation network a TU game will be defined, which takes into account the network topology, the weights associated with the arcs, and the demand based on the origin-destination matrix (weights associated with nodes). The nodes of the network represent the players of the TU game.

We aim at exploiting game-theoretic solution concepts developed during decades of research to identify the nodes that play a major role in the network. In particular, we shall use the so-called solution concept known as Shapley value [2], which represents a criterion according to which each node is attributed a value, in such a way that the larger the value the larger the node importance.  The Shapley value enjoys mathematical properties well-suited to the proposed analysis. Computational aspects related to the evaluation of the Shapley value will be investigated, too [6], studying the possibility of polynomial-time computation with respect to the network dimension.

Depending on whether the analysis focuses on the “physical nodes” or the “physical links”, the definition of the player changes. This research project considers both. When the transportation nodes (representing, e.g., intersections, transit terminals, bus stops, major points of interest, etc.) will be analyzed, the network on which the TU game will be defined is identical to the physical network. On the other hand, when arcs (e.g., road segments, transit routes, rail lines, etc.) will be analyzed, the network will be transformed in such a way that the physical links are modeled as nodes.

Link to the group/personal webpage:

http://www.dist.unige.it/msanguineti/

References

[1] S. Wasserman and K. Faust, Social Network Analysis: Methods and Applications. Vol. 8. Cambridge University Press, 1994.

[2] J. González-Díaz, I. García-Jurado, and M.G. Fiestras-Janeiro, An Introductory Course on Mathematical Game Theory. AMS, 2010.

[3] T.P. Michalak, Game-Theoretic Network Centrality – New Centrality Measures Based on Cooperative Game Theory, 2016. Available from: http://game-theoretic-centrality.com/index.html.

[4] Y . Hadas and, M. Sanguineti, An Approach to Transportation Network Analysis Via Transferable-Utility Games. 96th Annual Meeting of the Transportation Research Board, Transportation Research Board of the National Academies, Washington, DC, 8-12 gennaio 2017.

[5] Y. Hadas, G. Gnecco, M. Sanguineti, An Approach to Transportation Network Analysis Via Transferable    Utility

Title: Coupling of atmospheric and hydrological modelling