Estrazione automatica di informazioni da articoli scientifici in formato PDF e pubblicazione in Linked Open Data

La tesi ha lo scopo di introdurre Investiga, un'applicazione per l'estrazione automatica di informazioni da articoli scientifici in formato PDF e pubblicazione di queste informazioni secondo i principi e i formati Linked Open Data, creata per la tesi. Questa applicazione è basata sul Task 2 della SemPub 2016, una challenge che ha come scopo principale quello di migliorare l'estrazione di informazioni da articoli scientifici in formato PDF. Investiga estrae i capitoli di primo livello, le didascalie delle figure e delle tabelle da un dato articolo e crea un grafo delle informazioni così estratte collegate adeguatamente tra loro. La tesi inoltre analizza gli strumenti esistenti per l'estrazione automatica di informazioni da documenti PDF e dei loro limiti.

Trattamenti (dis)continui. Progetto di una nuova tipologia di macchina industriale per il trattamento superficiale continuo di prodotti farmaceutici.

Il progetto si concentra principalmente sullo studio formale e funzionale del carter di una nuova tipologia di macchina industriale per il trattamento superficiale (coating) continuo di prodotti famaceutici. I punti di interesse sono stati: la scelta del materiale, la forma del carter, l'interazione con l'utente, la visualizzazione della macchina finita, la proposta di un piano comunicativo e di una strategià d'identità del prodotto.

Deep-learning applicato all'identificazione automatica di frutta in immagini

Analisi delle fasi per la realizzazione di uno strumento di supporto gli agricoltori, dalla creazione di un dataset, all'addestramento e test di una rete neurale artificiale, con obiettivo la localizzazione del prodotto agricolo all'interno delle immagini.

Il fascicolo tecnico di una macchina astucciatrice: analisi dei rischi e calcoli strutturali di verifica richiesti dalle normative

La tesi, sviluppata presso l'azienda Universal-Pack, tratta del fascicolo tecnico di una macchina astucciatrice nelle sue varie parti: breve descrizione della macchina, analisi dei rischi secondo le normative collegate alla direttiva macchine, e calcoli strutturali di verifica di alcuni componenti, che sopportano sforzi alterni e conseguentemente vanno incontro a limiti di fatica.

Nonlinear control of the Reaction Wheel Pendulum

In this paper we introduce the Reaction Wheel Pendulum, a novel mechanical system consisting of a physical pendulum with a rotating bob. This system has several attractive features both from a pedagogical standpoint and from a research standpoint. From a pedagogical standpoint, the dynamics are the simplest among the various pendulum experiments available so that the system can be introduced to students earlier in their education. At the same time, the system is nonlinear and underactuated so that it can be used as a benchmark experiment to study recent advanced methodologies in nonlinear control, such as feedback linearization, passivity methods, backstepping and hybrid control. In this paper we discuss two control approaches for the problems of swingup and balance, namely, feedback linearization and passivity based control. We first show that the system is locally feedback linearizable by a local diffeomorphism in state space and nonlinear feedback. We compare the feedback linearization control with a linear pole-placement control for the problem of balancing the pendulum about the inverted position. For the swingup problem we discuss an energy approach based on collocated partial feedback linearization, and passivity of the resulting zero dynamics. A hybrid/switching control strategy is used to switch between the swingup and the balance control. Experimental results are presented.

Knowledge-Intensive Business Processes: Proceedings of the 1st International Workshop on Knowledge-intensive Business Processes [CEUR Workshop Proceedings, Volume 861]

Nowadays, Workflow Management Systems (WfMSs) and, more generally, Process Management Systems (PMPs) are process-aware Information Systems (PAISs), are widely used to support many human organizational activities, ranging from well-understood, relatively stable and structures processes (supply chain management, postal delivery tracking, etc.) to processes that are more complicated, less structured and may exhibit a high degree of variation (health-care, emergency management, etc.). Every aspect of a business process involves a certain amount of knowledge which may be complex depending on the domain of interest. The adequate representation of this knowledge is determined by the modeling language used. Some processes behave in a way that is well understood, predictable and repeatable: the tasks are clearly delineated and the control flow is straightforward. Recent discussions, however, illustrate the increasing demand for solutions for knowledge-intensive processes, where these characteristics are less applicable. The actors involved in the conduct of a knowledge-intensive process have to deal with a high degree of uncertainty. Tasks may be hard to perform and the order in which they need to be performed may be highly variable. Modeling knowledge-intensive processes can be complex as it may be hard to capture at design-time what knowledge is available at run-time. In realistic environments, for example, actors lack important knowledge at execution time or this knowledge can become obsolete as the process progresses. Even if each actor (at some point) has perfect knowledge of the world, it may not be certain of its beliefs at later points in time, since tasks by other actors may change the world without those changes being perceived. Typically, a knowledge-intensive process cannot be adequately modeled by classical, state of the art process/workflow modeling approaches. In some respect there is a lack of maturity when it comes to capturing the semantic aspects involved, both in terms of reasoning about them. The main focus of the 1st International Workshop on Knowledge-intensive Business processes (KiBP 2012) was investigating how techniques from different fields, such as Artificial Intelligence (AI), Knowledge Representation (KR), Business Process Management (BPM), Service Oriented Computing (SOC), etc., can be combined with the aim of improving the modeling and the enactment phases of a knowledge-intensive process. The 1st International Workshop on Knowledge-intensive Business process (KiBP 2012) was held as part of the program of the 2012 Knowledge Representation & Reasoning International Conference (KR 2012) in Rome, Italy, in June 2012. The workshop was hosted by the Dipartimento di Ingegneria Informatica, Automatica e Gestionale Antonio Ruberti of Sapienza Universita di Roma, with financial support of the University, through grant 2010-C26A107CN9 TESTMED, and the EU Commission through the projects FP7-25888 Greener Buildings and FP7-257899 Smart Vortex. This volume contains the 5 papers accepted and presented at the workshop. Each paper was reviewed by three members of the internationally renowned Program Committee. In addition, a further paper was invted for inclusion in the workshop proceedings and for presentation at the workshop. There were two keynote talks, one by Marlon Dumas (Institute of Computer Science, University of Tartu, Estonia) on "Integrated Data and Process Management: Finally?" and the other by Yves Lesperance (Department of Computer Science and Engineering, York University, Canada) on "A Logic-Based Approach to Business Processes Customization" completed the scientific program. We would like to thank all the Program Committee members for the valuable work in selecting the papers, Andrea Marrella for his valuable work as publication and publicity chair of the workshop, and Carola Aiello and the consulting agency Consulta Umbria for the organization of this successful event.

Practical stability of approximating discrete-time filters with respect to model mismatch

This paper establishes practical stability results for an important range of approximate discrete-time filtering problems involving mismatch between the true system and the approximating filter model. Practical stability is established in the sense of an asymptotic bound on the amount of bias introduced by the model approximation. Our analysis applies to a wide range of estimation problems and justifies the common practice of approximating intractable infinite dimensional nonlinear filters by simpler computationally tractable filters.

Dynamic positioning of marine craft using a port-Hamiltonian framework

Dynamic positioning of marine craft refers to the use of the propulsion system to regulate the vessel position and heading. This type of motion control is commonly used in the offshore industry for surface vessels, and it is also used for some underwater vehicles. In this paper, we use a port-Hamiltonian framework to design a novel nonlinear set-point-regulation controller with integral action. The controller handles input saturation and guarantees internal stability, rejection of unknown constant disturbances, and (integral-)input-to-state stability.

Reliable stabilization using a multi-controller configuration

Given a plant P, we consider the problem of designing a pair of controllers C1 and C2 such that their sum stabilizes P, and in addition, each of them also stabilizes P should the other one fail. This is referred to as the reliable stabilization problem. It is shown that every strongly stabilizable plant can be reliably stabilized; moreover, one of the two controllers can be specified arbitrarily, subject only to the constraint that it should be stable. The stabilization technique is extended to reliable regulation.

A cooperative game of a pair of learning automata

A cooperative game played in a sequential manner by a pair of learning automata is investigated in this paper. The automata operate in an unknown random environment which gives a common pay-off to the automata. Necessary and sufficient conditions on the functions in the reinforcement scheme are given for absolute monotonicity which enables the expected pay-off to be monotonically increasing in any arbitrary environment. As each participating automaton operates with no information regarding the other partner, the results of the paper are relevant to decentralized control.

Comments on a decentralized control of linear multivariable systems

A strongly connected decentralized control system may be made single channel controllable and observable with respect to any channel by decentralized feedbacks. It is noted here that the system example considered by Corfmat and Morse to illustrate this fact is already single channel controllable and observable, with respect to one of the channels. An alternate example which fits into the situation is presented in this item.

On the orthogonality of suboptimum estimate in the combined estimation and control problem

By deriving the equations for an error analysis of modeling inaccuracies for the combined estimation and control problem, it is shown that the optimum estimation error is orthogonal to the actual suboptimum estimate.

On the orthogonality of suboptimum estimate in the combined estimation and control problem

By deriving the equations for an error analysis of modeling inaccuracies for the combined estimation and control problem, it is shown that the optimum estimation error is orthogonal to the actual suboptimum estimate.

On the orthogonality of suboptimum estimate in the combined estimation and control problem

By deriving the equations for an error analysis of modeling inaccuracies for the combined estimation and control problem, it is shown that the optimum estimation error is orthogonal to the actual suboptimum estimate.

Natural actor-critic algorithms

We present four new reinforcement learning algorithms based on actor-critic, natural-gradient and functi approximation ideas,and we provide their convergence proofs. Actor-critic reinforcement learning methods are online approximations to policy iteration in which the value-function parameters are estimated using temporal difference learning and the policy parameters are updated by stochastic gradient descent. Methods based on policy gradients in this way are of special interest because of their compatibility with function-approximation methods, which are needed to handle large or infinite state spaces. The use of temporal difference learning in this way is of special interest because in many applications it dramatically reduces the variance of the gradient estimates. The use of the natural gradient is of interest because it can produce better conditioned parameterizations and has been shown to further reduce variance in some cases. Our results extend prior two-timescale convergence results for actor-critic methods by Konda and Tsitsiklis by using temporal difference learning in the actor and by incorporating natural gradients. Our results extend prior empirical studies of natural actor-critic methods by Peters, Vijayakumar and Schaal by providing the first convergence proofs and the first fully incremental algorithms.