ALMASat-1, ALMASat-EO and beyond: evolution of structural concepts and technologies towards multifunctional structures for microsatellites

Multifunctional Structures (MFS) represent one of the most promising disruptive technologies in the space industry. The possibility to merge spacecraft primary and secondary structures as well as attitude control, power management and onboard computing functions is expected to allow for mass, volume and integration effort savings. Additionally, this will bring the modular construction of spacecraft to a whole new level, by making the development and integration of spacecraft modules, or building blocks, leaner, reducing lead times from commissioning to launch from the current 3-6 years down to the order of 10 months, as foreseen by the latest Operationally Responsive Space (ORS) initiatives. Several basic functionalities have been integrated and tested in specimens of various natures over the last two decades. However, a more integrated, system-level approach was yet to be developed. The activity reported in this thesis was focused on the system-level approach to multifunctional structures for spacecraft, namely in the context of nano- and micro-satellites. This thesis documents the work undertaken in the context of the MFS program promoted by the European Space Agency under the Technology Readiness Program (TRP): a feasibility study, including specimens manufacturing and testing. The work sequence covered a state of the art review, with particular attention to traditional modular architectures implemented in ALMASat-1 and ALMASat-EO satellites, and requirements definition, followed by the development of a modular multi-purpose nano-spacecraft concept, and finally by the design, integration and testing of integrated MFS specimens. The approach for the integration of several critical functionalities into nano-spacecraft modules was validated and the overall performance of the system was verified through relevant functional and environmental testing at University of Bologna and University of Southampton laboratories.

Deep learning for computer vision: a comparison between convolutional neural networks and hierarchical temporal memories on object recognition tasks

In recent years, Deep Learning techniques have shown to perform well on a large variety of problems both in Computer Vision and Natural Language Processing, reaching and often surpassing the state of the art on many tasks. The rise of deep learning is also revolutionizing the entire field of Machine Learning and Pattern Recognition pushing forward the concepts of automatic feature extraction and unsupervised learning in general. However, despite the strong success both in science and business, deep learning has its own limitations. It is often questioned if such techniques are only some kind of brute-force statistical approaches and if they can only work in the context of High Performance Computing with tons of data. Another important question is whether they are really biologically inspired, as claimed in certain cases, and if they can scale well in terms of "intelligence". The dissertation is focused on trying to answer these key questions in the context of Computer Vision and, in particular, Object Recognition, a task that has been heavily revolutionized by recent advances in the field. Practically speaking, these answers are based on an exhaustive comparison between two, very different, deep learning techniques on the aforementioned task: Convolutional Neural Network (CNN) and Hierarchical Temporal memory (HTM). They stand for two different approaches and points of view within the big hat of deep learning and are the best choices to understand and point out strengths and weaknesses of each of them. CNN is considered one of the most classic and powerful supervised methods used today in machine learning and pattern recognition, especially in object recognition. CNNs are well received and accepted by the scientific community and are already deployed in large corporation like Google and Facebook for solving face recognition and image auto-tagging problems. HTM, on the other hand, is known as a new emerging paradigm and a new meanly-unsupervised method, that is more biologically inspired. It tries to gain more insights from the computational neuroscience community in order to incorporate concepts like time, context and attention during the learning process which are typical of the human brain. In the end, the thesis is supposed to prove that in certain cases, with a lower quantity of data, HTM can outperform CNN.

Performance studies of CMS workflows using Big Data technologies

Al Large Hadron Collider (LHC) ogni anno di acquisizione dati vengono raccolti più di 30 petabyte di dati dalle collisioni. Per processare questi dati è necessario produrre un grande volume di eventi simulati attraverso tecniche Monte Carlo. Inoltre l'analisi fisica richiede accesso giornaliero a formati di dati derivati per centinaia di utenti. La Worldwide LHC Computing GRID (WLCG) è una collaborazione interazionale di scienziati e centri di calcolo che ha affrontato le sfide tecnologiche di LHC, rendendone possibile il programma scientifico. Con il prosieguo dell'acquisizione dati e la recente approvazione di progetti ambiziosi come l'High-Luminosity LHC, si raggiungerà presto il limite delle attuali capacità di calcolo. Una delle chiavi per superare queste sfide nel prossimo decennio, anche alla luce delle ristrettezze economiche dalle varie funding agency nazionali, consiste nell'ottimizzare efficientemente l'uso delle risorse di calcolo a disposizione. Il lavoro mira a sviluppare e valutare strumenti per migliorare la comprensione di come vengono monitorati i dati sia di produzione che di analisi in CMS. Per questa ragione il lavoro è comprensivo di due parti. La prima, per quanto riguarda l'analisi distribuita, consiste nello sviluppo di uno strumento che consenta di analizzare velocemente i log file derivanti dalle sottomissioni di job terminati per consentire all'utente, alla sottomissione successiva, di sfruttare meglio le risorse di calcolo. La seconda parte, che riguarda il monitoring di jobs sia di produzione che di analisi, sfrutta tecnologie nel campo dei Big Data per un servizio di monitoring più efficiente e flessibile. Un aspetto degno di nota di tali miglioramenti è la possibilità di evitare un'elevato livello di aggregazione dei dati già in uno stadio iniziale, nonché di raccogliere dati di monitoring con una granularità elevata che tuttavia consenta riprocessamento successivo e aggregazione “on-demand”.