No description available.

Summary Driving error is a major factor in majority of all traffic accidents. To address this prob- lem, artificial cognitive systems based on human behaviour and inference of cognitive processes need to be developed for drivers safety and assistance in mundane scenarios. This research work aims to develop, analyse and demonstrate novel methodologies that can be useful for the design of cognitive driver assistance systems. In this thesis we propose to use a Perception-Action (P-A) learning approach to cognit- ive systems building for modelling human behaviour. The P-A approach seeks to reduce the complexity implicit within conventional environment/action-planning approaches, by mapping actions directly onto the resulting perceptual transitions, mitigating the use of intermediate representation and significantly reducing training requirements. As a prime strategy, we have decided to use machine learning techniques for building an adaptive system which provides descriptions of the driver's intentional behaviour using a psychological P-A model for human intentional modelling: Extended Control Model. Due to the (Highway Code) protocol based nature of the research problem, we pro- pose to use rule-based machine learning algorithms for characterising driver intentions (proof-of-concept model evaluation). A development and evaluation dataset comes from an instrumented car, comprising perceptual and driver-action/control inputs. For in-situ realtime applications, it is important that P-A modelling learning is con- sidered as an online problem. As such, we propose a novel methodology that utilizes a variational calculus approach to optimize an objective function defining system's pre- diction error, thus enabling P-A mapping to be treated as an online learning problem via gradient descent using partial derivatives. The proposed learning structure per- forms top-down modulation of low-level perceptual confidences (confidence functions of low-level input features) via the Jacobian of the higher levels of a Perception-Action hierarchy. Symbolic manipulation of perceptual confidences is carried using fuzzy-logic reasoning. A superior performance is achieved by the proposed learning framework compared to P-A learning without the top-down modulation. The approach developed also permits novel forms of context-dependant multi-level P-A mapping, important within the context of an intelligent driver assistance system. Key words: Perception-Action Modelling, Markov Logic Networks, Fuzzy Reasoning, Decision Tree Learning, First-order Logics, Variational Calculus.

No description available.

This research concerns issues arising in the design and implementation of advanced driver assistance systems, specifically longitudinal and lateral controllers; their effects on the dynamic performance of the vehicle, and their impact on the driver. The current state of the art is discussed as part of an extensive literature review, which highlights prominent gaps in the published research. There is a lack of understanding as to the effects of adverse environmental factors on the vehicle dynamics, and the effects of the systems' on the comfort of the driver. A novel twin track approach was taken to investigate these issues: the effects of the systems' on the vehicle dynamics were monitored using a range of off-line simulation tools, while the systems' impact on the driver was considered using an on-line driving simulator experiment. An adaptive cruise control system was developed, tuned to provide a comfortable response and implemented on a sophisticated off-line 9 degree of freedom vehicle model, with a non-linear tyre model. The system was tested in a range of environmental conditions. These simulations highlighted the good performance of the system in wet conditions, and revealed some possible driver conflicts. Two lateral control systems were developed, one based on a look down methodology, and the second on a more driver emulating look ahead approach. The systems were tested using the same high fidelity vehicle model, and an extensive range of suitable motorway manoeuvres. The systems were compared, proving the comfort and stability benefits of the look ahead system. The longitudinal and lateral control systems were integrated with the University of Leeds driving simulator. Ten subject drivers drove with and without the systems through a range of scenarios, some of which required evasive action. The impact of the systems on the driver, and the driver's response to safety critical scenarios was assessed. Results displayed little safety benefit of the systems in evasive scenarios, but drivers perceived improved awareness and comfort when under their control. The potential of advanced driver assistance systems to make driving a more comfortable and safe experience has been demonstrated, although the system engineer must consider the impact of the systems on the driver throughout their design and implementation.

A significant number of electronic control units (ECUs) are nowadays networked in automotive vehicles to help achieve advanced vehicle control and eliminate bulky electrical wiring. This, however, inevitably leads to increased complexity in vehicle fault diagnostics. Traditional off-board fault diagnostics and repair at service centres, by using only diagnostic trouble codes logged by conventional onboard diagnostics, can become unwieldy especially when dealing with intermittent faults in complex networked electronic systems. This can result in inaccurate and time consuming diagnostics due to lack of real-time fault information of the interaction among ECUs in the network-wide perspective. This thesis proposes a new framework for on-board knowledge-based diagnostics focusing on network level faults, and presents an implementation of a real-time in-vehicle network diagnostic system, using case-based reasoning. A newly developed fault detection technique and the results from several practical experiments with the diagnostic system using a network simulation tool, a hardware- in-the- loop simulator, a disturbance simulator, simulated ECUs and real ECUs networked on a test rig are also presented. The results show that the new vehicle diagnostics scheme, based on the proposed new framework, can provide more real-time network level diagnostic data, and more detailed and self-explanatory diagnostic outcomes. This new system can provide increased diagnostic capability when compared with conventional diagnostic methods in terms of detecting message communication faults. In particular, the underlying incipient network problems that are ignored by the conventional on-board diagnostics are picked up for thorough fault diagnostics and prognostics which can be carried out by a whole-vehicle fault management system, contributing to the further development of intelligent and fault-tolerant vehicles.

Le travail de la thèse s’articule sur le développement d’un nouveau type de capteurs de gaz à base des matériaux semi-conducteurs III-Nitrure (Les nitrures de gallium). Ces matériaux présentent de nombreux avantages qui pourraient être utilisées pour concevoir des capteurs NOx sensibles et sélectifs pour le contrôle des pollutions émises par la ligne d’échappement Diesel. Afin de limiter et déduire les gaz polluants émis par les moteurs à explosion en générale et les moteurs Diesel en particuliers (NO, NO2, NH3, CO, …), différentes normes européennes ont été établies. Pour respecter ces normes, plusieurs modifications sur les moteurs et les lignes d’échappement des véhicules ont été effectuées (filtres à particules, catalyseurs, capteurs NOx, …). Les capteurs NOx utilisés actuellement sont à base d’électrolyte solide. Ils sont basés dans leur fonctionnement sur la mesure de la concentration d’oxygène présente dans le gaz d’échappement qui permet de son tour l’estimation de la concentration totale des gaz NOx (mesure indirecte). Ces capteurs ne détectent pas le NH3 à la sortie de la ligne d’échappement, et ne donnent pas une information précise sur le rapport entre NO et NO2 (manque de sélectivité) qui est un facteur important pour le bon fonctionnement de catalyseur sélectif SCR (amélioration de rendement) ; d’où la nécessité d’un capteur de gaz plus performant et en particulier sélectif afin d’améliorer les systèmes de contrôle, de post-traitement et de diagnostic. Notre approche consiste à utiliser un transistor HEMT (High Electron Mobility Transistor) à gaz bidimensionnel d’électrons à base de nitrure de Gallium avec l’association d’une couche fonctionnelle à la place de la grille. L’interaction des molécules de gaz avec cette couche fonctionnelle donne une signature (variation de signal de sortie) spécifique pour chaque type de gaz qui aide à l’amélioration de la sélectivité. Le projet contient deux parties : l’optimisation de la structure choisie et l’optimisation de la couche fonctionnelle afin d’obtenir une détection sélective entre les différents gaz polluants. Cette technologie est intéressante pour développer des capteurs de gaz grâce aux possibilités de détecter des faibles variations de tensions et aux possibilités de fonctionnement dans des environnements sévères. La thèse de doctorat s’inscrit dans le cadre de l’OpenLab materials and processes en collaboration entre le laboratoire Georgia-Tech lorraine et l’entreprise Peugeot-Citroën PSA / The work of the thesis focuses on the development of a new type of gas sensors based III-Nitride semiconductor materials (gallium nitrides). These materials have many advantages that could be used to develop sensitive and selective NOx sensors for the control of pollution emitted by diesel exhaust line. To limit the polluting gases emitted by internal combustion engines in general and diesel in particular (NO, NO2, NH3, CO, ...), different European standards have been established. To meet these standards, anti-pollution systems (consisting of particle filters, catalysts, NOx sensors, ... etc) are used. NOx sensors currently used in automobiles are based on a solid electrolyte. Their operation is based on the measurement of the oxygen concentration. This enables an estimate of the total concentration of NOx gas (indirect measurement) after filtering NOx from O2 and decomposing NOx into O2. These sensors do not detect NH3 at the outlet of the exhaust line, and do not give accurate information on the relationship between NO and NO2 (lack of selectivity) which is important factor for an optimal functioning of selective catalyst (SCR performance improvement). Hence there exists a need for a more efficient and selective in particular gas sensor to improve the control systems, post-treatment and diagnosis. Our approach is to use a HEMT (High Electron Mobility Transistor) transistor based on gallium nitride with a combination of a functional layer instead of the gate. The interaction of the gas molecules with the functional layer gives a signature (output signal variation) specific for each type of gas that helps to improve the selectivity. The project contains two parts: the optimization of the chosen structure and the optimization of the functional layer in order to achieve selective detection between various gaseous pollutants. This technology is interesting for development of gas sensors through the possibility of detection low voltage variations and the possibility of operating in harsh environments. The thesis is part of OpenLab "Materials and Processes" in a collaboration between Georgia Tech-CNRS laboratory and the PSA Peugeot-Citroen Group

Suite à l’évolution de la réglementation relative aux dispositifs de retenue pour enfants en juillet 2013, une campagne d’évaluation expérimentale a permis de constater que la situation des plus jeunes impliqués dans un choc latéral était particulièrement critique. Pour y remédier, une méthodologie originale a été développée dans le but de dimensionner les matériaux d’absorption d’énergie « idéaux » à intégrer aux sièges-auto au niveau des zones d’impact de la tête, de l’épaule et du bassin. Une étude paramétrique basée sur ce concept a alors consisté à piloter la retenue de ces trois segments corporels dans le but de minimiser, soit les critères réglementaires (mannequins Q0 et Q1), soit les critères biomécaniques (modèle hybride Q0-6WOC). Les recommandations issues de cette étude ont ensuite été appliquées à un dispositif de retenue existant afin de l’optimiser en choc latéral. Après avoir vérifié que les valeurs d’accélération linéaire résultante de la tête et de HPC étaient bien inférieures aux seuils imposés par la réglementation, un modèle par Eléments Finis (EF) du siège-auto optimisé a été développé puis validé en situation dynamique. Finalement, après couplage de ce modèle EF du dispositif de retenue optimisé avec le modèle hybride Q0-6WOC, il a été montré que les résultats de simulation numérique en choc latéral étaient deux fois plus faibles que ceux obtenus avec le dispositif de retenue de référence. / Following changes in Child Restraint System (CRS) regulation in July 2013, an experimental CRS evaluation campaign has been conducted. Then, it has been noticed that the situation amongst the youngest involved in side impact was particularly critical. An original methodology has been therefore developed in order to define the “best” energy absorption materials to use in child seats for head, shoulder and pelvis areas. A parametric study in which the restraint of each body segments was independently piloted has then been conducted. The aim was to minimize either regulatory criteria of Q0 and Q1 dummies, or biomechanical criteria based on Q0-6WOC hybrid model. Those recommendations have thereafter been applied to improve an existing child restraint system under side impact. First, it has been checked that resultant linear acceleration of the head and HPC criteria were below regulation limits. Second, Finite Element (FE) model of the optimized car seat has been created and validated in several dynamical steps. Third, both reference child restraint system and optimized child seat has been evaluated in side impact with Q0-6WOC hybrid model. Finally, it has been showed that technical changes helped to cut in half the biomechanical results.