Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Mérite. --- Récompenses et punitions dans l'enseignement.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

The locus cœruleus (LC) is a small noradrenergic nucleus of high interest in neuroscience. As the main source of noradrenaline in the brain, it is involved in numerous cognitive functions such as arousal, attention and memory. In addition, previous studies have emphasised its relationship with the progression of neurodegenerative disorders such as Alzheimer's and Parkinson's diseases. The visualisation of the LC in magnetic resonance imaging (MRI) is realised through so-called neuromelanin-imaging in which magnetisation transfer (MT) effects are thought to be the main source of contrast. However, current researches insufficiently characterise the underlying contrast generation mechanisms given the poor understanding in the LC tissue properties. Efficiently visualising the locus coe ruleus is therefore a challenging process that requires further investigation.

The aim of this thesis is to provide an efficient tool for better understanding the underlying contrast mechanisms in the LC and thereby optimising its visualisation through multi-compartment spoiled gradient echo (SPGR) imaging. Firstly, the simulation of such a sequence through the extended phase graph with exchange (EPG-X) framework is performed in order to determine optimal sequence parameter values. After an experimentation phase, the outcomes are validated against the developed model. Simulation suggests that an optimal contrast can be achieved by maximising the power of the excitation pulse. It is expected that the use of optimal sequences compared to the one currently played out at the Wellcome Centre for Human Neuroimaging (WCHN) would improve the LC visualisation. Nonetheless, the lack of robust validation prevent the generalisation of these observations. Secondly, the SPGR model is extended by including a saturation pre-pulse for amplifying the MT effects. The contrast optimisation is realised through simulation according to the same formalism as previously while accounting for the MT pre-pulse. Optimal parameter values suggest that an improvement is achievable regarding the MT-weighted sequence currently played out through the maximisation of the power and the off-resonance frequency of the MT pulse as well as the time delay between the saturation and excitation pulses. Unfortunately, no validation was achievable for this configuration which should be investigated in future researches.

Because of the lack of knowledge about the contrast mechanisms in the LC, the optimisation of an MRI sequence effectively targeting the LC is complex. Therefore, due to the numerous parameters involved and the poor confidence regarding their impact, future studies should focus on a better characterisation of this structure and the inherent contrast mechanisms.

MRI --- Locus cœruleus --- Magnetisation transfer --- SPGR --- Contrast --- Ingénierie, informatique & technologie > Multidisciplinaire, généralités & autres

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

This thesis consists in developing a position sensor in order to control a perma-
nent magnet synchronous motor (PMSM). It is the aim of an intership in the
MITIS company. The company is developing micro-CHP (Combined Heat and
Power) systems. These systems are mainly composed of a PMSM, a compressor,
a combustor and a turbine. In fact, the open-loop control already implemented
presents different issues of operating that need to be solved. For that, a closed-
loop control is necessary for a better efficiency of the global system. This one
needs an angular position sensor for its feedback information that is developed
in this work. The choice of the sensor is an Eddy current sensor characterized
for its low maintenance, its accuracy and its large range of rotation speed mea-
surement. Indeed, for this project, a rotation speed of 120 000 rpm need to be
reached by the motor. Moreover, it does not need a proximity with the magnet.
This thesis is composed of different chapters. First, the concept of micro tur-
bine, PMSM and its controller are presented. Then, a chapter that explain in
details how the sensor had been designed and manufactured regarding the me-
chanical and manufactured constraints. After, another chapter is dedicated to
the software development of the sensor with the difficulties encountered such as
the rapidity of data acquisition. Finally, a chapter about different closed-loop
controls that could be implemented is established and the advantages and dis-
advantages for each type of control are supported by simulations.
Finally, some good results are obtained. The sensor is unfortunately not tested
on a real motor but an experiment that allows to reach 5000 rpm is performed
and is successful. Moreover, the choice of the speed and torque closed-loop con-
trol is very promising regarding the results of the simulations with a very high
speed reached (120 000 rpm) and a low consumption of the current in the stator
windings (about 30A).

motor --- PMSM --- permanent --- magnet --- angle --- speed --- closed-loop --- rotation --- micro-turbine --- Ingénierie, informatique & technologie > Ingénierie électrique & électronique

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Machine learning has proven itself to be a useful tool in many scientific fields. Among the plethora of applications, modern Robotics aims to incorporate artificial intelligence, in order to tackle tasks that were previously unthinkable to be performed by a robot. Grasping is a classic example of such a challenge. Being able to grasp and manipulate a wide range of objects reliably represents a major stepping stone in both productivity and flexibility for modern robots. 
The first step of grasping an object is to identify it correctly. To that end, most state of the art approaches to robotic grasping include all kinds of visual input to facilitate the location and identification of the target object. This work aims to investigate a different approach, where object classification is tackled using exclusively the robot's joint information. More specifically this project explores the question: "How useful is robot joint information for the classification of objects?" The main contributions of this work are various benchmark experiments as a proof of concept, additional methods aiming to increase classification performance, as well as the application on a set of real world objects. The results and observations of the different problems settings in this work lead to the conclusion that joint information is indeed useful and exploitable, but also comes with clear limitations and challenges, some of which are easier to circumvent than others.

Grasping, Deep Learning, Robotics, Machine Learning --- Ingénierie, informatique & technologie > Sciences informatiques