An Embedded In-Cabin Lightweight High-Performance 3D Gaze Estimation System

Gaze estimation has gained interest in recent years for being an important cue to obtain information about the internal cognitive state of humans. Regardless of whether it is the 3D gaze vector or the point of gaze (PoG), gaze estimation has been applied in various fields, such as: human robot interaction, augmented reality, medicine, aviation and automotive. In the latter field, as part of Advanced Driver-Assistance Systems (ADAS), it allows the development of cutting-edge systems capable of mitigating road accidents by monitoring driver distraction. Gaze estimation can be also used to enhance the driving experience, for instance, autonomous driving. It also can improve comfort with augmented reality components capable of being commanded by the driver's eyes. Although, several high-performance real-time inference works already exist, just a few are capable of working with only a RGB camera on computationally constrained devices, such as a microcontroller. This work aims to develop a low-cost, efficient and high-performance embedded system capable of estimating the driver's gaze using deep learning and a RGB camera. The proposed system has achieved near-SOTA performances with about 90% less memory footprint. The capabilities to generalize in unseen environments have been evaluated through a live demonstration, where high performance and near real-time inference were obtained using a webcam and a Raspberry Pi4.

Web Component per un Design System: un caso di studio aziendale

I contenuti e i servizi offerti dal Web hanno subito negli anni una costante e continua evoluzione, dovuti alla maggiore disponibilità di dispositivi in grado di navigarlo. Oggi i requisiti e le aspettative che gli utenti hanno nei confronti delle applicazioni Web sono sempre maggiori, desiderano un accesso ai contenuti sempre più rapido, interfacce semplici e facili da usare, oltre che reattive, e che tali contenuti siano accessibili da una vasta gamma di dispositivi che presentino sempre più funzionalità. Le aziende devono essere pronte a rispondere a queste esigenze e a fornire agli utenti finali la miglior esperienza possibile, rimanendo aggiornati sulle tecnologie per la creazione di applicazioni Web. Questo è ancora più importante per un'azienda possiede più prodotti, sviluppati da team diversi che usano tecnologie diverse. Per alcune aziende è importante che i propri prodotti, sebbene trattino tematiche differenti, si presentino con interfacce che rimandino al proprio marchio, non solo grazie al nome o al logo, quanto più nei componenti utilizzati per creare le interfacce. Succede così che i vari team devono progettare e sviluppare i componenti nella propria tecnologia, in modo che abbiano le stesse funzionalità, stesso stile e stesso comportamento in ogni situazione. Il più delle volte questo è difficile da realizzare e anche costoso da mantenere. Riuscire a centralizzare lo sviluppo di questi elementi in un unico punto aiuta l'azienda a mantenere bassi i costi di manutenzione e a rendere omogenea l'esperienza degli utenti tra i vari prodotti. Obiettivo del lavoro svolto è illustrare le potenzialità e l'utilità fornite dall'introduzione di una suite di componenti personalizzati, seguendo lo standard dei Web Component, all'interno dei prodotti forniti da una grande impresa. L'analisi si concentra sulll'esperienza di chi utilizza tali componenti all'interno dei propri progetti per creare l'interfaccia utente da presentare poi agli utenti finali.

Cure Induced Residual Stresses on Thick Composite Components Manufactured by Additive Process

This study investigates the effect of an additive process in manufacturing of thick composites. Airstone 780 E epoxy resin and 785H Hardener system is used in the analysis since it is widely used wind turbine blade, namely thick components. As a fiber, fabric by SAERTEX (812 g/m2) with a 0-90 degrees layup direction is used. Temperature overshoot is a major issue during the manufacturing of thick composites. A high temperature overshoot leads to an increase in residual stresses. These residual stresses are causing warping, delamination, dimensional instability, and undesired distortion of composite structures. A coupled thermo-mechanical model capable of predicting cure induced residual stresses have been built using the commercial FE software Abaqus®. The possibility of building thick composite components by means of adding a finite number of sub-laminates has been investigated. The results have been compared against components manufactured following a standard route. The influence of pre-curing of the sub-laminates has also been addressed and results compared with standard practice. As a result of the study, it is found that introducing additive process can prevent temperature overshoot to occur and benefits the residual stresses generation during the curing process. However, the process time required increases by 50%, therefore increasing the manufacturing costs. An optimized cure cycle is required to minimize process time and cure induced defects simultaneously.

Study of an advanced transport system exploiting magnetic levitation: the Beckhoff XPlanar

The transport system is one of the most important components to be chosen in the design of an automatic machine. There is a wide variety of different choices that can be made in picking this element, each one having its own strengths and its own drawbacks. If it is desired to obtain some elaborate behaviour from the transport system, it is a good idea to think about some flexible and advanced solutions. Among these transport systems, the newest is the Beckhoff XPlanar. This transport system exploits magnetic levitation to move some passive magnetic movers on a completely customizable plane, in an entirely contact-free way. This provides a fast, clean, and noiseless motion, which is extremely desirable in a modern automatic machine. The purpose of this Thesis is to analyse the potentialities and the problems of this new device, starting from the basics. After having presented in detail the topic, an analysis on the hardware components needed to build this system is performed. Then, it is conducted a study on the concepts needed to know how to build a controller having the purpose of dealing with this system. After that, the various types of motion are studied and executed and, later on, some experiments on the real kit are carried out. These studies start from the diagnostic and involve other analyses that are used to test the limits of this transport system. In performing these analyses, it is noticed how the kit presents some problems in reaching the limits of the dynamics. Finally, two different types of station cycle are implemented, which are useful to get a rough idea on the potentialities of this new advanced transport system.

Design and implementation of a vibrational condition monitoring system

Industrial companies, particularly those with induction motors and gearboxes as integral components of their systems, are utilizing Condition Monitoring (CM) systems more frequently in order to discover the need for maintenance in advance, as traditional maintenance only performs tasks when a failure has been identified. Utilizing a CM system is essential to boost productivity and minimize long-term failures that result in financial loss. The more exact and practical the CM system, the better the data analysis, which adds to a more precise maintenance forecast. This thesis project is a cooperation with PEI Vibration Monitoring s.r.l. to design and construct a low-cost vibrational condition monitoring system to check the health of induction motors and gearboxes automatically. Moreover, according to the company's request, such a system should have specs comparable to NI 9234, one of the company's standard Data Acquisition (DAQ) boards, but at a significantly cheaper price. Additionally, PEI VM Company has supplied all hardware and electronic components. The suggested CM system is capable of highprecision autonomous monitoring of induction motors and gearboxes, and it consists of a Raspberry Pi 3B and MCC 172 DAQ board.