Pauline Michel

Ms. Pauline Michel

University of Paris-Saclay, Gif-sur-Yvette, Île-de-France, France
orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 August 2022 in Electronics
Verified
Journal Article
Open access
PDF file
PMPauline Michel
Pauline Michel
SBSamir Bouaziz
Samir Bouaziz
FDFlavien Delgehier
Flavien Delgehier
SEStéphane Espié
Stéphane Espié
Human-in-the-loop driving simulation aims to create the illusion of driving by stimulating the driver’s sensory systems in as realistic conditions as possible. However, driving simulators can only produce a subset of the sensory stimuli that would be available in a real driving situation, depending on the degree of refinement of their design. This subset must be carefully chosen because it is crucial for human acceptability. Our focus is the design of a physical dynamic (i.e., motion-based) motorcycle-riding simulator. For its instrumentation, we focused on the rider acceptability of all sub-systems and the simulator as a whole. The significance of our work lies in this particular approach; the acceptability of the riding illusion for the rider is critical for the validity of any results acquired using a simulator. In this article, we detail the design of the hardware/software architecture of our simulator under this constraint; sensors, actuators, and dataflows allow us to (1) capture the rider’s actions in real-time; (2) render the motorcycle’s behavior to the rider; and (3) measure and study rider/simulated motorcycle interactions. We believe our methodology could be adopted by future designers of motorcycle-riding simulators and other human-in-the-loop simulators to improve their rendering (including motion) quality and acceptability.

18 October 2024 in Data in Brief
Verified
Journal Article
Open access
Bastien VinckeBastien Vincke
Bastien Vincke
PMPauline Michel
Pauline Michel
Abdelhafid El OuardiAbdelhafid El Ouardi
Abdelhafid El Ouardi
BLBruno Larnaudie
Bruno Larnaudie
FDFlavien Delgehier
Flavien Delgehier
ABAbderrahmane Boubezoul
Abderrahmane Boubezoul
The dynamics of a two-wheeled vehicle represent a higher level of complexity, with a high variety of rider practices, peculiarly during the bend-taking maneuver. This complexity is one reason why research on motorcyclists’ practiced trajectories is limited compared to cars. Another reason is the low representation of motorcycles in the land motor vehicle fleet.

13 December 2024 in ITM Web of Conferences
Verified
Journal Article
Open access
PDF file
SRSaša Radosavljevic
Saša Radosavljevic
ARAlain Rivero
Alain Rivero
Sergio Rodriguez FlorezSergio Rodriguez Florez
Sergio Rodriguez Florez
Abdelhafid El OuardiAbdelhafid El Ouardi
Abdelhafid El Ouardi
PMPauline Michel
Pauline Michel
PVPhilippe Vanheeghe
Philippe Vanheeghe
SNCF Réseau introduces a novel multi-modal embedded monitoring system, addressing challenges in railway infrastructure maintenance. The design incorporates visual, inertial, and sound sensors, enhancing adaptability, improving overall detection precision, and could reduce operational costs. This study addresses visual defects detection that can be integrated in a multi-modal monitoring system. The paper details the system’s architecture, synchronisation methods, and decision fusion process to improve the precision of limited mono-modal systems. A deep-learning visual based railway defects inspection was explored. Results show that small CNN (Yolov8 nano) can achieve similar (Yolov8 XL) high precision (mAP@0.5 ≥ 0.89) for a small number of objects (9) while improving implementation capability on embedded systems.

7 April 2026
Verified
Thesis
Closed access
PMPauline Michel
Pauline Michel
Le « mal du simulateur », qui s’apparente jusqu’à un certain point au mal des transports, est particulièrement problématique. La compréhension des causes du mal du simulateur et l'identification de procédés pour réduire l’occurrence de ce dernier sont des enjeux importants dans la mesure où elle introduit un biais double dans les résultats acquis : la population étudiée est filtrée « à la porte » et où le conducteur qui souffre de mal du simulateur risque d’avoir des comportements de conduite altérés.L'hypothèse que nous défendons dans le cadre de cette thèse concerne la relation entre « contrôlabilité » du système (le simulateur) et mal du simulateur. En simulation de conduite, le conducteur fait partie intégrante de la boucle : il initie les mouvements et reçoit les réponses du système. Notre hypothèse est que, dans ce cadre, un mauvais contrôle du déplacement du véhicule simulé induit des déplacements erratiques, perçus par le conducteur comme ne correspondant pas à ses attentes (du fait de son expérience de conduite antérieure) et donc non-acceptables.Pour tester cette hypothèse, nous avons repensé les modalités d’interfaçage conducteur/simulateur d’un simulateur de conduite moto préexistant : architecture matérielle et logicielle, structure capteurs/actionneurs de la colonne de direction, modèle de véhicule implémenté. Le simulateur, dans sa version actuelle, permet un meilleur degré de contrôlabilité par le conducteur pour des vitesses de déplacement moyennes à hautes.

13 December 2024 in ITM Web of Conferences
Verified
Journal Article
Open access
PDF file
FGFatima Zahra Guerrouj
Fatima Zahra Guerrouj
Sergio Rodriguez FlorezSergio Rodriguez Florez
Sergio Rodriguez Florez
Abdelhafid El OuardiAbdelhafid El Ouardi
Abdelhafid El Ouardi
PMPauline Michel
Pauline Michel
Belkacem Ould-BouamamaBelkacem Ould-Bouamama
Belkacem Ould-Bouamama
Real-time object detection is crucial for autonomous vehicles, and YOLO (You Only Look Once) algorithms have demonstrated their effectiveness for this purpose. This study examines the performance of YOLOv4 [3] for real-time object detection on an embedded architecture. We focus on optimizing the computationally intensive convolution operations by employing the cuDNN library to achieve efficient inference. The evaluation assesses critical performance metrics, including object detection accuracy in terms of Mean Average Precision (mAP) and inference latency on the embedded architecture. We conduct a comparative analysis using the publicly available KITTI [7] database. The reported results establish a benchmark between the parallelized YOLOv4 model and the baseline implementation, assessing the advantages of cuDNN acceleration for real-time object detection on resource-constrained devices.