Menu Fermer

SéminDocs

Qu’est ce que sont les SeminDocs : Les SeminDocs sont des exposés fait par des doctorants ou jeunes chercheurs travaillant ou de passage au LIRMM. Ils ont pour but de favoriser les échanges scientifiques entre les doctorants des différents départements. Le conseil des doctorants organise régulièrement des sessions durant l’année. Il s’agit de venir écouter des présentations de vulgarisation assez courtes (entre 10 et 20 minutes chacune) de camarades doctorants venus parler de leur travail ou de sujets de recherche qui les intéressent. Vu la diversité des disciplines s’exerçant au LIRMM, l’objectif est de rendre ces présentations  accessibles à tous et le but premier est surtout de satisfaire une curiosité et ouverture scientifiques vers d’autres domaines de recherche.

Mercredi 26 février 2020

1st talk : Nicolas Hlad (INFO)
Software product line

In this semindoc, we will present the engineering of software product line. This engineering is applied for the conception, development and maintenance of large set of products family.

2nd talk :  Rodolfo Villalobos Martinez (ROB)
Modeling of energy consumption for autonomous underwater vehicle (AUV)

During an exploration mission of an Autonomous Underwater Vehicle (AUV) the energy available in the battery is limited. Therefore the extension of the mission depends directly from the power consumption of the AUV. In order to plan the mission we propose a model of energy consumption for the full system. The model is divided in propulsive energy (speed controller, motor, propeller) and ancillary energy (software and sensors), this will allow to have an accurate energy consumption estimation prediction for different exploration missions.

Mercredi 24 avril 2019

1er exposé : Lucas Lavenir (ROB)
Guidage du geste chirurgical pour la pose de neuroprothèses auditives par fusion d’images échographique / tomodensitométrique

Les implants cochléaires sont les premières neuroprothèses fonctionnelles à avoir été développées pour les sujets humains. Elles restaurent la perception auditive chez les individus atteints de surdité sévère à profonde. L’implantation de ces prothèses requiert l’insertion d’un porte-électrode dans la cochlée, permettant ainsi la stimulation électrique de l’extrémité des fibres du nerf auditif. Aujourd’hui,l’insertion de ce porte-électrode est une étape critique de la procédure chirurgicale : les chirurgiens ne peuvent contrôler visuellement la trajectoire du porte-électrode et doivent s’en remettre au retour tactile uniquement. L’absence de contrôle visuel est à l’origine de nombreux dommages causés aux structures cochléaires et la perte consécutive de l’ouïe résiduelle. Un tel risque empêche l’implantation d’implant cochléaire chez les individus atteints de surdité plus légère. Ce travail de thèse vise à développer un nouvel outil permettant de guider l’insertion du porte-électrode cochléaire grâce à la fusion d’images d’échographie préopératoire et de tomodensitométrie préopératoire.

2ème exposé : Tom Davot (INFO)
Méthodes informatiques pour la production de génome après le séquençage de l’ADN

Le séquençage de l’ADN est une opération permettant de déterminer la séquence de nucléotides dans un génome. Cependant cette opération nécessite de couper ce génome en petits fragments, nécessitant d’être assemblés pour pouvoir recréer la séquence complète. Le nombre de fragments étant considérable (de l’ordre de quelques milliers à plusieurs millions), il est très difficile de reconstituer la séquence complète à la main, et des outils et méthodes informatiques sont nécessaires pour venir en aide aux biologistes afin d’avoir une vision globale du génome séquencé. Nous verrons les différentes étapes de cette reconstitution et comment les outils utilisés en informatique sont mis en oeuvre pour produire la séquence d’ADN.

3ème exposé : Safa Mhamdi (MIC)
Towards Improvement of Mission Mode Failure Diagnosis for System-on-Chip

Abstract: In critical (e.g. automotive) applications, Systems-on-Chip (SoC) failures that occurred during mission mode (in the field) are the most critical since they may lead to catastrophic effects. In this context, diagnosis is crucial in order to establish the root cause of observed failures with the best accuracy. With the advent of very deep submicron technologies (i.e. 7 nm), achieving such level of accuracy will become more and more difficult with today’s intra-cell diagnosis tools based on effect- cause or cause-effect paradigms. This will compromise the success of subsequent Physical Failure Analysis (PFA) done on defective SoCs. Machine Learning (ML) is now used in numerous classification problems where the knowledge on some data can be used to classify a new instance of such data. In particular, several ML-based solutions exist to address volume diagnosis for yield improvement. These learning-guided diagnosis approaches start from an existing set of defect candidates and try to minimize this set (eliminate bad candidates) owing to the use of ML tools and numerous data collected during production test (e.g. thousands of failed chips with candidates correctly labeled). Although efficient in volume diagnosis, these approaches cannot be used to identify the root cause of failures in customer returns, since only one failed chip is investigated in this case, with no information about the defective behavior of some other similar chips used in the same conditions (environment, workload, etc.). In this paper, we propose a new learning-guided approach for diagnosis of mission mode failures in customer returns. The proposed approach directly produces a minimum set of good candidates derived from the application of the learning-guided intra-cell diagnosis flow. Results obtained on a set of benchmark circuits, and comparison with a commercial intra-cell diagnosis tool, show the feasibility, effectiveness and accuracy of the proposed approach.

Mardi 18 Juin 2019

1er exposé :  Philippe Lambert (ROB)
Toward autonomous fault tolerant mission with performance guarantee.

Based on a decade of research results, this talk will explain the followed trajectory of the EXPLORE team to construct and manage fully autonomous missions. Firstly, the experimental context is presented justifying the need of complex mission development, autonomy, fault tolerance and performance guarantee. This need of autonomy is analyzed according to the view-point of Cyber-Physical and Autonomic Systems development goals and challenges. Then we summarize the main characteristics the PANORAMA (Performance and AutoNOmy using Resources Allocation MAnagement) approach that has been developed to guarantee efficient robotic mission using behavioral autonomy management.

2ème exposé : Iago Bonnici (INFO)
Effects of Input Addition in Learning for Adaptive Games: Towards Learning with Structural Changes

Adaptive Games (AG) involve a controller agent that continuously feeds from player actions and game state to tweak a set of game parameters in order to maintain or achieve an objective function such as the flow measure defined by Mihály Csíkszentmihályi (Hungarian-American psychologist). This can be considered a Reinforcement Learning (RL) situation, so that classical Machine Learning (ML) approaches can be used. On the other hand, many games naturally exhibit an incremental gameplay where new actions and elements are introduced or removed progressively to enhance player’s learning curve or to introduce variety within the game. This makes the RL situation unusual because the controller agent input/output signature can change over the course of learning. In this paper, we get interested in this unusual “protean” learning situation (PL). In particular, we assess how the learner can rely on its past shapes and experience to keep improving among signature changes without needing to restart the learning from scratch on each change. We first develop a formal model of the PL problem. Then, we address two first elementary signature changes: “input addition”, and “input deletion” with Recurrent Neural Networks (RNNs) in an idealized PL situation. As a first result, we find that it is possible to benefit from prior learning in RNNs even if the past controller agent signature had different inputs. The use of PL in AG thus remains encouraged. Investigating output/feedback addition/deletion, and translating these results to generic PL will be part of future works.

3ème exposé : Clément Touzet (MIC)
In-Memory computing for cryptographic applications

With the increasing requirements in energy efficient and high-performance computing solutions, In-Memory Computing (IMC) has been presented as a promising solution to reduce the power consumption due to data transfer between processing and memory units. In this presentation, we’ll see the state-of-the-art of IMC with a special emphasis on the different memory technologies (SRAM, ReRAM, and MRAM) that can be used to build an IMC architecture. Then, we will look at cryptographic operation that can be implmented in the IMC architecture.

Lundi 04 novembre 2019

1er exposé : Gabriel Volte (INFO)
The Differential Harvest Problem.

 The Differential Harvest Problem consists in optimizing the harvesting time while harvesting at least a given quantity of grapes of the best quality. It can be seen as an enhanced capacity Vehicle Routing Problem with business constraints. The Differential Harvest Problem consists in optimizing harvests of different grape qualities in vineyards so that we obtain a certain quantity, denoted by $R_min$, of good quality grapes. There are two types of grape quality: A quality and B quality grapes. Without loss of generality, let us assume that A grapes quality is of better quality than B grapes quality. Thanks to agronomic information obtained a priori, it is possible to map a vineyard ready for harvesting by distinguishing areas according to the quality of the grapes. There are geolocalized harvesting machines equipped with 2 hoppers (harvest tanks), with a capacity of $C_max$, capable of using such a map to store the best quality grapes in one hopper, and the other grapes in the second. When one of the two hoppers is full, both must be emptied into a bin located at the edge of the plot.

2ème exposé : Quentin Massone (ROB)
Underwater environment 3D mapping : application to karst networks.

My talk will explain the followed trajectory of the EXPLORE team in the exploration/inspection of underwater confined environment. As generic case studies, the EXPLORE project focuses on karst network exploration which are underwater galleries. This is a major and urgent issue for public authorities concerned by the prospection, protection and management of the groundwater resource in karst regions. My work consist in mapping these water galleries through the use of a stereo camera. So, I will present the principles of stereo vision and the constraints related to the environment. I will conclude this talk with a first approach I am implementing which use 2 cameras and the halo of an artificial light.

Mercredi 04 décembre 2019 à 10h

1er exposé : Vincent Iampietro (INFO)
Introduction to formal methods applied to software engineering.

During this presentation, I will give an overview of formal methods applied to software engineering. In particular, we will explore the perks and drawbacks related to the application of such methods. Eventually, we will focus on a specific kind of formal methods called “deductive methods”, which I will illustrate leveraging the Coq proof assistant.

2ème exposé : José Luis Vilchis Medina (ROB)
Modeling of Resilient Systems in Non-monotonic Logic.

In this talk I will present a resilient model to pilot an aircraft based on a non-monotonic logic. This model is capable of handling solutions from incomplete, contradictory information and exceptions. This is a very well known problem in the field of AI, which has been studied for more than 40 years. Default logic is used to formalise situations and then find possible actions. Thanks to this logic, we can transform the piloting rules to defaults. The control of the system is done via the property of resilience, I redefine this property as the integration of the non-monotonic logic in the Minsky model.

3ème exposé : Thibault Vayssade (MIC)
Test of IOT transceiver from digital test equipment.

The testing cost of transceiver dedicated to the Internet Of Thing (IOT) is very high due to the use of extremely expensive and specifics analog equipment. This talk will explain the operation of an IOT transceiver and the test challenge of this products. Then a possible solution based on the use of a digital test equipment and a processing algorithm will be introduced.