Structure learning of graphical models for task-oriented robot grasping

In the collective imaginaries a robot is a human like machine as any androids in science fiction. However the type of robots that you will encounter most frequently are machinery that do work that is too dangerous, boring or onerous. Most of the robots in the world are of this type. They can be found in auto, medical, manufacturing and space industries. Therefore a robot is a system that contains sensors, control systems, manipulators, power supplies and software all working together to perform a task. The development and use of such a system is an active area of research and one of the main problems is the development of interaction skills with the surrounding environment, which include the ability to grasp objects. To perform this task the robot needs to sense the environment and acquire the object informations, physical attributes that may influence a grasp. Humans can solve this grasping problem easily due to their past experiences, that is why many researchers are approaching it from a machine learning perspective finding grasp of an object using information of already known objects. But humans can select the best grasp amongst a vast repertoire not only considering the physical attributes of the object to grasp but even to obtain a certain effect. This is why in our case the study in the area of robot manipulation is focused on grasping and integrating symbolic tasks with data gained through sensors. The learning model is based on Bayesian Network to encode the statistical dependencies between the data collected by the sensors and the symbolic task. This data representation has several advantages. It allows to take into account the uncertainty of the real world, allowing to deal with sensor noise, encodes notion of causality and provides an unified network for learning. Since the network is actually implemented and based on the human expert knowledge, it is very interesting to implement an automated method to learn the structure as in the future more tasks and object features can be introduced and a complex network design based only on human expert knowledge can become unreliable. Since structure learning algorithms presents some weaknesses, the goal of this thesis is to analyze real data used in the network modeled by the human expert, implement a feasible structure learning approach and compare the results with the network designed by the expert in order to possibly enhance it.

Toward a peripheral view of manufacturing networks

This doctoral thesis unfolds into a collection of three distinct articles that share an interest in supply firms, or “peripheral firms”. The three studies offer a novel theoretical perspective that I call the peripheral view of manufacturing networks. Building on the relational view literature, this new perspective identifies the supplier-based theoretical standpoint to analyze and explain the antecedents of relational rents in manufacturing networks. The first article, the namesake of the dissertation, is a theoretical contribution that explains the foundations of the “peripheral view of manufacturing networks”. The second article “Framing The Strategic Peripheries: A Novel Typology of Suppliers” is an empirical study with the aim to offer an interpretation of peripheries’ characteristics and dynamics. The third article, “What is Behind Absorptive Capacity? Dispelling the Opacity of R&D” presents an example of general theory development by using data from peripheral firms.

Structural design, manufacturing and testing of the new wing for the CSIR's Modular UAS in composite materials

In the framework of an international collaboration with South Africa CSIR, the structural design, manufacturing and testing of the new wing for the Modular UAS in composite materials has been performed.

Hot wire manufacturing and resolution effects in high Reynolds number flows

Lo studio della turbolenza è di fondamentale importanza non solo per la fluidodinamica teorica ma anche perchè viene riscontrata in una moltitudine di problemi di interesse ingegneristico. All'aumentare del numero di Reynolds, le scale caratteristiche tendono a ridurre le loro dimensioni assolute. Nella fluidodinamica sperimentale già da lungo tempo si è affermata l'anemometria a filo caldo, grazie ad ottime caratteristiche di risoluzione spaziale e temporale. Questa tecnica, caratterizzata da un basso costo e da una relativa semplicità, rende possibile la realizzazione di sensori di tipo artigianale, che hanno il vantaggio di poter essere relizzati in dimensioni inferiori. Nonostante l'ottima risoluzione spaziale degli hot-wire, infatti, si può verificare, ad alto numero di Reynolds, che le dimensioni dell'elemento sensibile siano superiori a quelle delle piccole scale. Questo impedisce al sensore di risolvere correttamente le strutture più piccole. Per questa tesi di laurea è stato allestito un laboratorio per la costruzione di sensori a filo caldo con filo di platino. Sono in questo modo stati realizzati diversi sensori dalle dimensioni caratteristiche inferiori a quelle dei sensori disponibili commercialmente. I sensori ottenuti sono quindi stati testati in un getto turbolento, dapprima confrontandone la risposta con un sensore di tipo commerciale, per verificarne il corretto funzionamento. In seguito si sono eseguite misure più specifiche e limitate ad alcune particolari zone all'interno del campo di moto, dove è probabile riscontrare effetti di risoluzione spaziale. Sono stati analizzati gli effetti della dimensione fisica del sensore sui momenti statistici centrali, sugli spettri di velocità e sulle funzioni di densità di probabilità.

Manufacturing and characterization of amorphous silicon alloys passivation layers for silicon hetero-junction solar cells

Nel presente lavoro di tesi magistrale sono stati depositati e caratterizzati film sottili (circa 10 nm) di silicio amorfo idrogenato (a-Si:H), studiando in particolare leghe a basso contenuto di ossigeno e carbonio. Tali layer andranno ad essere implementati come strati di passivazione per wafer di Si monocristallino in celle solari ad eterogiunzione HIT (heterojunctions with intrinsic thin layer), con le quali recentemente è stato raggiunto il record di efficienza pari a 24.7% . La deposizione è avvenuta mediante PECVD (plasma enhanced chemical vapour deposition). Tecniche di spettroscopia ottica, come FT-IR (Fourier transform infrared spectroscopy) e SE (spettroscopic ellipsometry) sono state utilizzate per analizzare le configurazioni di legami eteronucleari (Si-H, Si-O, Si-C) e le proprietà strutturali dei film sottili: un nuovo metodo è stato implementato per calcolare i contenuti atomici di H, O e C da misure ottiche. In tal modo è stato possibile osservare come una bassa incorporazione (< 10%) di ossigeno e carbonio sia sufficiente ad aumentare la porosità ed il grado di disordine a lungo raggio del materiale: relativamente a quest’ultimo aspetto, è stata sviluppata una nuova tecnica per determinare dagli spettri ellisometrici l’energia di Urbach, che esprime la coda esponenziale interna al gap in semiconduttori amorfi e fornisce una stima degli stati elettronici in presenza di disordine reticolare. Nella seconda parte della tesi sono stati sviluppati esperimenti di annealing isocrono, in modo da studiare i processi di cristallizzazione e di effusione dell’idrogeno, correlandoli con la degradazione delle proprietà optoelettroniche. L’analisi dei differenti risultati ottenuti studiando queste particolari leghe (a-SiOx e a-SiCy) ha permesso di concludere che solo con una bassa percentuale di ossigeno o carbonio, i.e. < 3.5 %, è possibile migliorare la risposta termica dello specifico layer, ritardando i fenomeni di degradazione di circa 50°C.

Discrete Event Systems based Design Patterns for Diagnosability Analysis of Automated Manufacturing Systems

The main goal of this thesis is to facilitate the process of industrial automated systems development applying formal methods to ensure the reliability of systems. A new formulation of distributed diagnosability problem in terms of Discrete Event Systems theory and automata framework is presented, which is then used to enforce the desired property of the system, rather then just verifying it. This approach tackles the state explosion problem with modeling patterns and new algorithms, aimed for verification of diagnosability property in the context of the distributed diagnosability problem. The concepts are validated with a newly developed software tool.

Tecniche di fault tolerance in swarm robotics

Questa tesi presenta e discute le sfide per ottenere sistemi di swarm robotis affidabili e tolleranti ai guasti e quindi anche alcuni metodi per rilevare anomalie in essi, in modo tale che ipotetiche procedure per il recupero possano essere affrontate, viene sottolineata inoltre l’ importanza di un’ analisi qualitativa dei guasti.

Additive manufacturing: studio delle non conformità geometriche e dimensionali in componenti realizzati mediante tecnica fused deposition modelling

La fabbricazione additiva è una classe di metodi di fabbricazione in cui il componente viene costruito aggiungendo strati di materiale l’uno sull’altro, sino alla completa realizzazione dello stesso. Si tratta di un principio di fabbricazione sostanzialmente differente da quelli tradizionali attualmente utilizzati, che si avvalgono di utensili per sottrarre materiale da un semilavorato, sino a conferire all’oggetto la forma desiderata, mentre i processi additivi non richiedono l’utilizzo di utensili. Il termine più comunemente utilizzato per la fabbricazione additiva è prototipazione rapida. Il termine “prototipazione”’ viene utilizzato in quanto i processi additivi sono stati utilizzati inizialmente solo per la produzione di prototipi, tuttavia con l’evoluzione delle tecnologie additive questi processi sono sempre più in grado di realizzare componenti di elevata complessità risultando competitivi anche per volumi di produzione medio-alti. Il termine “rapida” viene invece utilizzato in quanto i processi additivi vengono eseguiti molto più velocemente rispetto ai processi di produzione convenzionali. La fabbricazione additiva offre diversi vantaggi dal punto di vista di: • velocità: questi processi “rapidi” hanno brevi tempi di fabbricazione. • realizzazione di parti complesse: con i processi additivi, la complessità del componente ha uno scarso effetto sui tempi di costruzione, contrariamente a quanto avviene nei processi tradizionali dove la realizzazione di parti complesse può richiedere anche settimane. • materiali: la fabbricazione additiva è caratterizzata dalla vasta gamma di materiali che può utilizzare per la costruzione di pezzi. Inoltre, in alcuni processi si possono costruire pezzi le cui parti sono di materiali diversi. • produzioni a basso volume: molti processi tradizionali non sono convenienti per le produzioni a basso volume a causa degli alti costi iniziali dovuti alla lavorazione con utensili e tempi di setup lunghi.

Influence of Surface Roughness on Mechanical Properties of Two Computer-aided Design/Computer-aided Manufacturing (CAD/CAM) Ceramic Materials

SUMMARY The aim of this study was to evaluate the influence of surface roughness on surface hardness (Vickers; VHN), elastic modulus (EM), and flexural strength (FLS) of two computer-aided design/computer-aided manufacturing (CAD/CAM) ceramic materials. One hundred sixty-two samples of VITABLOCS Mark II (VMII) and 162 samples of IPS Empress CAD (IPS) were ground according to six standardized protocols producing decreasing surface roughnesses (n=27/group): grinding with 1) silicon carbide (SiC) paper #80, 2) SiC paper #120, 3) SiC paper #220, 4) SiC paper #320, 5) SiC paper #500, and 6) SiC paper #1000. Surface roughness (Ra/Rz) was measured with a surface roughness meter, VHN and EM with a hardness indentation device, and FLS with a three-point bending test. To test for a correlation between surface roughness (Ra/Rz) and VHN, EM, or FLS, Spearman rank correlation coefficients were calculated. The decrease in surface roughness led to an increase in VHN from (VMII/IPS; medians) 263.7/256.5 VHN to 646.8/601.5 VHN, an increase in EM from 45.4/41.0 GPa to 66.8/58.4 GPa, and an increase in FLS from 49.5/44.3 MPa to 73.0/97.2 MPa. For both ceramic materials, Spearman rank correlation coefficients showed a strong negative correlation between surface roughness (Ra/Rz) and VHN or EM and a moderate negative correlation between Ra/Rz and FLS. In conclusion, a decrease in surface roughness generally improved the mechanical properties of the CAD/CAM ceramic materials tested. However, FLS was less influenced by surface roughness than expected.

Robotics in ear, nose and throat (ENT) surgery

The three main classes of robotics in ear, nose and throat (ENT) surgery (telemanipulation, image-guided functional servoing, and computer numerical control) are discussed and important examples of applications are described to show both technological and clinical developments. As access to many anatomical features of the head requires very small and compact tools to accurately perform procedures, examples are give in ear surgery where manipulation of the minute ossicles requires fine, dexterous movements.

Prion removal capacity of plasma protein manufacturing processes

BACKGROUND The variant Creutzfeldt-Jakob disease incidence peaked a decade ago and has since declined. Based on epidemiologic evidence, the causative agent, pathogenic prion, has not constituted a tangible contamination threat to large-scale manufacturing of human plasma-derived proteins. Nonetheless, manufacturers have studied the prion removal capabilities of various manufacturing steps to better understand product safety. Collectively analyzing the results could reveal experimental reproducibility and detect trends and mechanisms driving prion removal. STUDY DESIGN AND METHODS Plasma Protein Therapeutics Association member companies collected more than 200 prion removal studies on plasma protein manufacturing steps, including precipitation, adsorption, chromatography, and filtration, as well as combined steps. The studies used a range of model spiking agents and bench-scale process replicas. The results were grouped based on key manufacturing variables to identify factors impacting removal. The log reduction values of a group are presented for comparison. RESULTS Overall prion removal capacities evaluated by independent groups were in good agreement. The removal capacity evaluated using biochemical assays was consistent with prion infectivity removal measured by animal bioassays. Similar reduction values were observed for a given step using various spiking agents, except highly purified prion protein in some circumstances. Comparison between combined and single-step studies revealed complementary or overlapping removal mechanisms. Steps with high removal capacities represent the conditions where the physiochemical differences between prions and therapeutic proteins are most significant. CONCLUSION The results support the intrinsic ability of certain plasma protein manufacturing steps to remove prions in case of an unlikely contamination, providing a safeguard to products.

Good manufacturing practice-compliant validation and preparation of BM cells for the therapy of acute myocardial infarction

BACKGROUND: Intracoronary application of BM-derived cells for the treatment of acute myocardial infarction (AMI) is currently being studied intensively. Simultaneously, strict legal requirements surround the production of cells for clinical studies. Thus good manufacturing practice (GMP)-compliant collection and preparation of BM for patients with AMI was established by the Cytonet group. METHODS: As well as fulfillment of standard GMP requirements, including a manufacturing license, validation of the preparation process and the final product was performed. Whole blood (n=6) and BM (n=3) validation samples were processed under GMP conditions by gelafundin or hydroxyethylstarch sedimentation in order to reduce erythrocytes/platelets and volume and to achieve specifications defined in advance. Special attention was paid to the free potassium (<6 mmol/L), some rheologically relevant cellular characteristics (hematocrit <0.45, platelets <450 x 10(6)/mL) and the sterility of the final product. RESULTS: The data were reviewed and GMP compliance was confirmed by the German authorities (Paul-Ehrlich Institute). Forty-five BM cell preparations for clinical use were carried out following the validated methodology and standards. Additionally three selections of CD34+ BM cells for infusion were performed. All specification limits were met. Discussion In conclusion, preparation of BM cells for intracoronary application is feasible under GMP conditions. As the results of sterility testing may not be available at the time of intracoronary application, the highest possible standards to avoid bacterial and other contaminations have to be applied. The increased expense of the GMP-compliant process can be justified by higher safety for patients and better control of the final product.