PhD : selection by topics
Mathematics - Numerical analysis - Simulation
Biochemistry
Optics - Laser optics - Applied optics
Solid state physics, surfaces and interfaces
Biotechnologies,nanobiology
Biotechnology, biophotonics
Analytic chemistry
Electronics and microelectronics - Optoelectronics
Materials and applications
Thermal energy, combustion, flows
Mechanics, energetics, process engineering
Physical chemistry and electrochemistry
Medical imaging
Instrumentation
Electromagnetism - Electrical engineering
Radiation-matter interactions
Chemistry
Automatics, Remote handling
Nuclear physics
Ultra-divided matter, Physical sciences for materials
Neutronics
Biotechnologies,nanobiology
Biotechnology, biophotonics
Biochemistry
Materials and applications
Solid state physics, surfaces and interfaces
Electronics and microelectronics - Optoelectronics
Theoretical Physics
Ultra-divided matter, Physical sciences for materials
Chemistry
Solid state physics, chemistry and nanosciences >> Mathematics - Numerical analysis - Simulation
14 propositions.
Radiolocation Algorithms Adapted to High-Precision Ultra Wideband Rake Receivers Using Multiple Antennas
Making possible unprecedented user-centric services (monitoring/smart inventory of personal goods from a smartphone, context-aware indoor navigation, etc.), new radiolocation capabilities appeared in wireless networks are on the verge of modifying in depth our daily life and habits. This revolution might even be comparable to that of GSM cellular communications or GPS-based car navigation in the past. In this context, an Impulse Radio ? Ultra Wideband receiver (IR-UWB) endowed with multiple fingers (RAKE) has been recently integrated with success. The latter claims low consumption (a few 10s of mW in average), while enabling the detection of weak incoming signals and enjoying scalable data rates (<50Mbp) that depends on the targeted transmission ranges (up to several 100s of meters). One more particularity lies in the capacity to estimate precisely the arrival time (TOA) of transmitted pulses, without any high-speed sampling operation on the receiver side. This solution can thus theoretically provide peer-to-peer distance estimates between radio devices within a precision of a few centimetres. Using multiple antennas on both transmitter and/or receiver sides can also help to improve signal sensitivity ahead (and hence, transmission ranges accordingly), as well as to jointly deliver angles of departure/arrival to complement estimated arrival times. However, in most of indoor non-controlled operating conditions (radio device hold in the user's hand or in the pocket), propagation phenomena can play a harmful role on the final localization performance. Practically, the obstruction of radio links by bodies and/or walls, or even an unfavourable device attitude, can lead to significant errors on measured distances and finally on position estimates. But the time/space correlation of received multipath components, as well as the highly timely and localized nature of the most severe obstruction cases, can be beneficial to try to enhance performance. In this PhD work, the main contribution consists in theoretically investigating, designing and evaluating new radiolocation algorithms that could take profits from the specificities of such high-precision IR-UWB devices. In particular, significant multipath components resolved at the receiver fingers shall be exploited further. We will also consider deploying multiple antennas at fixed reference nodes (base stations), as well as at mobile terminals. One goal would be to guarantee performances equivalent to that of optimistic visibility cases. One side objective would be to provide new detection functions (obstructions, wrong device attitude or geometric ambiguities).
See the summary of the offerSub-nanometer electron tomography for semiconductor devices
FDSOI technology aims at reaching 28nm CMOS transistors by limiting leakage currents and device complexity. It requires a 3D characterisation featuring sub-nanometer resolution. At the moment, no technique allows to perform such a characterization. Still in the semiconductor field, intensive research is carried out on quantum dots, for ultimate transistor and optical applications. The lateral localisation of these quantum dots is today impossible due to the lack of 3D nanoimaging tools. Electron tomography is a 3D imaging technique that allows reconstructing a 3D image of an object from its 2D projections. So far, resolutions achievable are in the nanometer range. This research topic deals with the development of electron tomography in order to reach a resolution of a few Angstroms. Specifically, electron microscopy in STEM mode, that today allows observing atomic columns, will be extended in 3D so that the whole 3D distribution of atoms will be retrieved with an isotropic, deep sub-nanometre resolution. Preliminary works have already been done by few teams in the world. The research work will consist in developing the technique by tuning the electron optics, developing new 3D and efficient sampling and using the last generation of reconstruction algorithms. Of course, the work will be applied to the most recent devices, after properly shaping the samples using ion beam sample preparation.
See the summary of the offerNew graph partitioning models for tasks assignment and routing on massively parallel architectures
The emergence, in recent years, of the so-called many-core processor architectures, that is microprocessors integrating hundreds if not thousands cores on a single chip, has created a new field of applications for graph partitioning problems. Indeed, graph partitioning issues are central when one needs to map a network of tasks on such an architecture. Still, the first models which have been studied only coarsely reflect the true complexity of the underlying architectures and new, more complete, models are required. In this context, a first objective of the proposed PhD thesis will be to focus on defining new models which will then induce new variants of the graph partitioning problem. A second objective will be to focus on studying their mathematical structure so as to propose, and experimentally validate, practically relevant resolution algorithms.
See the summary of the offerCorrection of scatter radiation in multi-energy radiographic and tomographic imaging
X-ray imaging is a well-known technique used in the medical diagnostic domain, testing of industrial samples, and security. Conventional X-ray detectors provide a signal proportional to the energy of photons deposited into the detector, integrated on the total energy range. The pixel value is in fact the sum of two signals: the primary flux one, coming linearly from the detector, measuring the photons not attenuated by the photoelectric effect, and the scattered radiation, the interacted photons being modified in direction and energy. The scattered radiation induces modifications of image values, thus quantification bias, and is significant as soon as 2D detectors are used. Numerous methods exist for scatter correction. The recently emerged semiconductor detectors, providing the full energetic spectra for each pixel, imply the evolution of these processing methods. The global objective of the thesis is to develop estimation and correction methods of the scattered radiation when using spectral detectors. Numerical simulation and experimental devices will be used. Generalisation of the existing mono-spectral methods will be tested, then original methods based on all energy bins will be investigated. Both radiographic and tomographic modes will be considered.
See the summary of the offerLearning for tracking: exploiting context and probability reasoning for reinforcing knowledge.
The Laboratory of Vision and Content Engineering (LVIC) brings together 80 researchers around research topics which are the analysis of multimedia data, the augmented reality and the scene analysis. Through these activities, the LVIC works with large groups such as Renault, Thales, SNCF as well as with a network of innovative SMEs DiotaSoft, Arcure ... Through R & D projects, the laboratory develops automatic video analysis systems in real time integrating advanced techniques of objects tracking in the scene. In this thesis, the objective of the laboratory is to segment and track pedestrians in real urban scenes which induces dealing with highly variable appearance and frequent interactions. A strong scientific problem lies in the ability to learn a relevant visual signature of the persons (that is to say, a discriminating and robust one), to achieve an effective tracking function. To do this, one must dispose at the same time of a large number of descriptions as well as accurate and qualitative data. Today, 3D tracking approaches of multiple agents are based on rigid models of shapes and appearances and make use of advanced modules background segmentation, classification and appearance's modeling. These approaches provide a large number of reliable descriptions. In contrast, many recent approaches, concerning the monitoring of an unknown target from an onboard camera, exploit 2D flexible and deformable targets models based on interests points and involve highly dynamic appearance modeling modules. They seem more relevant to obtain accurate and qualitative data. Finaly, the objective is to bring together the best concepts of these two areas. The essence of the thesis is the exploitation of context and probabilistic reasoning to increase knowledge about the appearance of people and be able to distinguish them at all times.
See the summary of the offerNew medical imaging systems exploiting the energy dispersive X-ray diffraction with spectrometric CdZnTe based detector.
The context of the thesis is the measurement of energy dispersive X-rays diffraction (EDXRD) using spectrometric semiconductor based detectors for medical applications. Today, most of medical imaging systems exploit the X-ray radiation transmitted through tissues. The recent emergence of energy resolved detectors, (semiconductors based in particular) allows considering the exploitation of another type of radiation: the diffracted radiation (or small angle scattered). Measurement of diffracted spectra can provide new information related to the molecular structure of the examined tissues, thus improving the final diagnosis. The modalities can be the detection and analysis of tumors for breast imaging, of kidney stones, bone densitometry, or even the diffraction measurement of nanoparticles used as contrast agents. Spectrometric CdZnTe based detectors, developed in the LDET laboratory, combine performance (energy resolution, sensitivity, spatial resolution, compactness) necessary to measure relevant diffracted spectra in the medical context under clinical condition. The aim of the thesis is therefore to study the feasibility of such a detection system. It will be a global approach of system conception, combining geometric design, and adaptation of technology and architecture of spectrometric detectors.
See the summary of the offerUsing tactile information in robotic command - Application to dexterous manipulation
In the rapidly growing field of robotic, catching and manipulating without breaking it are difficult tasks on which the CEA is working for many years. With this aim, the LIST has developed an artificial hand inspired from human one. This artificial hand has 5 fingers, 24 articular degrees of freedom and an optimal control. On the other hand, using MEMS technologies the LETI has developed 3 axis force sensors which size and sensibility allow to use them as tactile sensors. Moreover, we have demonstrated that they have interesting properties for surface analysis and more particularly for friction coefficient estimation, which is involved in sliding. The association of these two technologies allows intending to handle in an optimal way the manipulation of slippery objects. With this objective, the thesis will first have to bring solutions for relevant feature extraction from the sensors signals. Then, fusion of these parameters with those of the hand and of the object to be manipulated will be considered. Finally the command of the hand will have to be adapted to take into account all the available information. The work will be realized partially in the LETI and partially in the LIST. The integration of the sensors into the hand will be realized. Experiments will be conducted and the results will have to be interpreted. Finally, command algorithms will be developed, and will be validated experimentally.
See the summary of the offerDesign, Fabrication and Test of o a 1Tbps-optoelectronic Transceiver for very short reach transmissions
Wavelength Division Multiplexing techniques, and more recently, Multi-level signaling formats have established themselves in optical data transmissions over long distances. Implemented with standalone, big-footprint-components, the techniques are energyvorous, and they can't address yet, shorter reach applications, despite the need for higher transmission rates. CMOS photonics (compatible with the low-cost-fabrication technology of CMOS-electronics) is an emerging technology allowing the integration of optical and analog/digital electronic functions within a unique module, through 3D CMOS processes. Lower cost, lower footprint, and lower power solutions are expected for short reach data-transmissions ranging from rack-to-rack, board-to-board and module-to-module (e.g.: switch). The targeted applications are High Performance Computers, and Data centers. In order to establish itself as the technology of choice for such applications, Silicon Photonics must demonstrate its increased power efficiency as compared with the existing technologies, while showing its ability for higher transmission data rates. Relying on the Silicon photonic components developed in the lab (lasers, (de)multiplexers, photo-detectors, ?), and making use of the electronic drivers developed by our partner at Caltech, the doctoral fellow will propose and develop innovative solutions for the transmission and reception of 1Tbps-data, implementing different multiplexing and modulation techniques. The main activities will consist in - design of transmitters and receivers scalable to 1Tbps, and modeling of the optical transmission - implementation of the solutions (circuit design including layouts according to the Design Rule Manual of the CMOS fabrication) - characterization of the optical transmission implementing the developed transceivers (emitter/receiver)
See the summary of the offerPerformance modeling of PEMFC coupling discrete and continuous approaches.
The objective of this thesis is to develop a new generation of more realistic performance model of PEM Fuel Cells based on Pore Network Modeling (PNM) that allow 1) to have a better determination of fluid transport (gas, vapor, liquid) in the MEA (Membrane Electrode Assembly) which is crucial for the performance and durability 2) to better link the structures of the components to their performance. This work will be based on the original PNM models developed in the laboratory for the study of transfers in the diffusion layers (gas and liquid) and in the active layers (gas, liquid, heat, charge and electrochemistry), by direct coupling or by homogenization. An experimental validation will be done on tests performed in parallel. Improvements of components (pore size distribution, wettability, material compounds distribution...) will be discussed and proposed. This thesis will be done in close collaboration between the CEA / LCPEM recognized as a specialist in component development and modeling of Fuel Cells, and CNRS/IMFT recognized as a specialist in the development of the approach pore networks in porous media.
See the summary of the offerMulti-level modeling for hybrid energetic systems including renewable generation: Combining electric and thermal modeling for energy management
The large scale integration of renewable energy generation into power systems may be challenging due to the variability and limited predictability of the resource. In order to reduce such impacts, renewable sources may be combined into hybrid power systems, which include controllable thermal and electric units. The main objective of the present work is to develop multi-energy models for the energy management of such hybrid systems. The first step consists in combining thermal and electric models for the energy management of hybrid systems, such as a building including photovoltaic generation. These energetic models are used to compute the system optimal control, which defines optimal setpoints for the controllable units, from forecasts of the renewable generation and of the consumption. The second step consists in combining the given energetic models with physical models for the simulation of the system operation. These models will inform about the component operation simulation based on the setpoints which are given by the energy management models. Also, these physical models will be used to validate the energy management models and algorithms in order to deploy them for real-time operation.
See the summary of the offerDevelopment of algorithms for the Tracking of Optimum Point of Charge for electrochemical storage technologies. Application to Li-Ion and ZEBRA batteries chargers
The fast charge algorithms of electrochemical batteries decrease inevitably the global charge/discharge efficiency. The previous studies held by the CEA-Tech have permitted to determine a parameter for setting the charging rate at an optimum value that can reduce the charging time while minimizing the losses. This optimal rate called OPC (Optimum Point of Charge) varies dynamically with the states of the battery. This thesis aims to extend these initial studies operated on Lithium-Ion cells to storage systems of Li-Ion packs and High temperature Sodium batteries (ZEBRA) including reversible inverters. The purpose of this thesis is the development and the implementation of OPC tracking algorithms into the inverters.
See the summary of the offerInteractive analysis and validation of multi-agent based simulations with fuzzy logic
Multi-agent based simulations (MABS) have been widely used in complex systems modeling like in socio-spatial (urban dynamics), sociotechnical (smart-grids) or biological phenomena. These simulations can be very complex to analyze when there is a large number of agents, behaviors and interactions involved. However, only few tools enable to analyze and validate these simulations, none of them enabling an analysis guided by the user. This thesis aims at tackling this issue by proposing new analysis and validation tools in order to ease the understanding of the underlying phenomena. These tools will be based on the use of fuzzy logic that permits to get intelligible results automatically or interactively, as well as visualization tools that enable to get quick and intuitive analysis results of one or several simulation. The definition of fuzzy concepts, that gives sense to numerical values, as well as fuzzy rules, making causal relations, will be used to analyze the behavior of homogeneous groups of agents identified by clustering in the simulation, test hypotheses for a specific scenario and search the configurations that meet user or expert requirements.
See the summary of the offerExperimental and numerical study of 3D metallic powder densification using Hot Isostatic Pressing
Main objectives of this PhD work is to study powder mechanical behavior at low (particle reorganization) and high temperature (viscoplasticity) of a metallic powder, with the aim of the modelling of a complex 3D parts compaction by Finite Element Method. The PhD will contain an experimental part (instrumented HIP experiments, traction and compaction experiments,SEM TEM XRD observations, tomography or/and FIB), a numerical part (Discret Element Method, Finite element method, ODE resolution) and finally a more applied part consisting in a complex part compaction simulation using new modeles developped in this three years work.
See the summary of the offerWireless and Privacy: Device Free Motion Detection into the wild
Wireless networks are nowadays deployed everywhere. If their primary goals is to transmit data, it seems possible to subvert radio waves for other purposes such as motion tracking or localization. During this thesis, we want to investigate how intrusive are the wireless indicators like the Received Signal Strength Indicator (RSSI) and Link Quality Indicator (LQI). How much information on a person can be inferred from these indicators ?
See the summary of the offer