Search

National audience; Cet article présente les notions de base d'autonomie opérationnelle et décisionnelle des véhicules intelligents évoluant sur des routes ouvertes à la circulation publique. Il décrit les principaux systèmes nécessaires à la navigation autonome de ces véhicules : représentation de l'espace, localisation, cartes, perception, décision, planification et contrôle. La plupart des notions sont liées au domaine de la robotique mobile, au contrôle et aux systèmes autonomes. Elles sont décrites d'une façon générale et des exemples concrets les illustrent. Un lecteur novice dans ce domaine pourra ainsi appréhender les différentes facettes de cette problématique complexe. Nous donnons également des éléments bibliographiques qui permettront au lecteur d'aller plus loin en fonction de ses centres d'intérêt.

International audience; To navigate autonomously, a vehicle must be able to localize itself with respect to its driving environment and the vehicles with which it interacts. This work presents a decentralized cooperative localization method. It is based on the exchange of Local Dynamic Maps (LDM), which are cyberphysical representations of the physical driving environment containing poses and kinematic information about nearby vehicles. An LDM acts as an abstraction layer that makes the cooperation framework sensor-agnostic, and it can even improve the localization of a sensorless communicating vehicle. With this goal in mind, this work focuses on the property of consistency in LDM estimates. Uncertainty in the estimates needs to be properly modeled, so that the estimation error can be statistically bounded for a given confidence level. To obtain a consistent system, we first introduce a decentralized fusion framework that can cope with LDMs whose errors have an unknown degree of correlation. Second, we present a consistent method for estimating the relative pose between vehicles, using a 2D LiDAR with a point-to-line metric within an iterative-closest-point approach, combined with communicated polygonal shape models. Finally, we add a bias estimator in order to reduce position errors when non-differential GNSS receivers are used, based on visual observations of features geo-referenced in a High-Definition (HD) map. Real experiments were conducted, and the consistency of our approach was demonstrated on a platooning scenario using two experimental vehicles. The full experimental dataset used in this work is publicly available.; Pour naviguer de manière autonome, un véhicule doit être capable de se localiser par rapport à son environnement et par rapport aux véhicules avec lesquels il interagit. Ce travail présente une méthode de localisation coopérative décentralisée. Il est basé sur l'échange de cartes locales dynamiques (CLD), qui sont des représentations cyberphysiques de l'environnement de conduite physique contenant des poses et des informations cinématiques sur les véhicules à proximité. Une CLD agit comme une couche d'abstraction qui rend la coopération indépendante des capteurs. Elle peut de plus améliorer la localisation d'un véhicule communicant sans capteur. Avec cet objectif en tête, ce travail se concentre sur la consistence des estimations des CLD. L'incertitude dans les estimations doit être correctement modélisée, afin que l'erreur d'estimation puisse être statistiquement limitée pour un niveau de confiance donné. Pour obtenir un système consistant, nous introduisons d'abord une fusion décentralisée qui peut faire face aux CLD dont les erreurs ont un degré de corrélation inconnu. Ensuite, nous présentons une méthode consistante pour estimer la pose relative entre les véhicules, en utilisant un LiDAR 2D avec une méthode ICP (Iterative Closest Point) basée sur une correspondance point à ligne, combinée à des modèles polygonaux communiqués. Enfin, nous ajoutons un estimateur de biais afin de réduire les erreurs de position lorsque des récepteurs GNSS non différentiels sont utilisés, sur la base d'observations visuelles de marquages géoréférencées dans une carte haute définition (HD). Des expériences réelles ont été menées, et la consistence de notre approche a été démontrée sur un scénario de conduite en convoi utilisant deux véhicules expérimentaux. L'ensemble des données expérimentales utilisées dans ce travail a été rendu public.

International audience

International audience

International audience

International audience; — Localization is an important issue for autonomous vehicle navigation. In this paper, we present a one-dimensional cooperative localization problem by using curvilinear abscissa with respect to a reference map. This is made possible by using a high-definition map containing accurate information, such as lane markings which can be used to reduce significantly the cross-track and heading errors. In the context of cooperative localization, a vehicle can use its perception capabilities along with V2V communication to retrieve the relative pose of another vehicle in order to improve the accuracy of its own pose estimate. Experimental results with non independent data coming from communication devices show the necessity to use a more consistent algorithm for data fusion than the classical Kalman filter, such as covariance intersection. The results presented in this paper also show that the one-dimensional cooperative localization algorithm is particularly useful on curved roads since the accuracy increases when cooperative vehicles drive in such conditions.

International audience; Accurate localization with high availability is a key requirement for autonomous vehicles. It remains a major challenge when using automotive sensors such as single-frequency GNSS receivers, a lane detection camera and proprioceptive sensors. This paper describes a method that enables the estimation of standalone L1-GNSS errors by integrating the measurements from a forward looking camera matched with lane markings stored in a digital map. It includes a parameter identification method for a shaping model which is evaluated using experimental data. An algebraic observability study is then conducted to prove that the proposed state vector is fully observable in a road-oriented frame. This observability property is the basis to develop a road-centred Extended Kalman Filter (EKF) which can maintain the observability of every component of the state vector on any road, whatever its orientation. To accomplish this the flter needs to handle road changes, which it does using bijective transformations. The fllter was implemented and tested intensely on an experimental vehicle for driverless valet parking services. Field results have shown that the performance of the estimation process is better than solutions based on EKF implemented in a fixed workingframe. The proposed filter guarantees that the drift along the road direction remains bounded. This is very important when the vehicle navigates autonomously. Further, the road-centred modeling improves the accuracy, consistency and robustness of the localization solver.

International audience; Map-matching can be used to estimate the Horizontal Uncertainty Level (HUL) of GNSS position fixes. Integrity monitoring is indeed an important issue for autonomous vehicles navigation. The method is based on the use of a high definition map that stores accurate information about theroad network. This additional source of information is crucial for autonomous navigation. The matched position is computed using proprioceptive sensors from the car and GNSS fixes that are handled using a precautionary principle with Horizontal Protection Levels (HPL). A Particle Filter is used for its ability to manage multiple hypotheses if needed. Estimating different likely map-matched hypotheses allows to determine the levelof uncertainty of the GNSS which is defined as the maximum distance between a map-matched hypothesis and a given GNSS position. This distance can be seen as a Map-Aided Horizontal Uncertainty Level (MA-HUL), providing a confidence indicator to the vehicle for integrity monitoring. This paper presents the multi-hypotheses map-matching algorithm and a method to compute the MA-HUL values in real-time. Experimental resultscarried out in open road conditions support the evaluation and show that this metric provides reliable confidence information.

National audience; —Occupancy grids are popular in autonomous navigation for encoding obstacle information into grid cells to provide real-time environmental models. However, very few studies have been carried out on encoding lane and traffic information in grids. This information refines the world model up to the lane level which is important in many situations to enable vehicles to follow basic road rules, such as lane keeping or lane changes in case of overtaking for instance. Usual approaches consist in detecting lane boundaries using on-board cameras or lidars but the problem is tricky when the road is multi-lanes or in challenging weather conditions. In this work, we propose to tackle this problem by using a vectorial prior map that stores detailed lane level information. We take advantage of the pose estimation from a localization solver and propagate the estimation uncertainty over the grids cells. Both Bayesian and Evidential models are presented and some of their special characteristics are highlighted and compared. Real results carried on public roads with the same real-time software are reported to support the comparison.

International audience; Detection and identification of faulty measurements like multipath is of prime importance for radio-navigation. A method based on interval analysis and constraint propagation is presented in this paper and applied to GPS positioning. This method is able to detect and identify erroneous measurements in real-time simultaneously with the computation of the positioning domain. With bounded-error measurements (GPS pseudoranges on visible satellites here), fault-detection is guaranteed not to trigger false alarms. A q-relaxed robust estimation method is studied. By the use of a breadth-first exploration strategy and of measurement consistency counters, the algorithm can promptly signal a detected fault during the set-inversion process. It can also be stopped at any time of the evaluation, and can instantly return the solution subpaving and the list of identified erroneous measurements. We demonstrate than the method is able to handle simultaneous faults as long as they are fewer than the degree of relaxation. The method has been evaluated with real GPS pseudodistances in an urban environment with a low-cost high-sensitivity GPS receiver providing numerous faulty multipath measurements and we report experiments. As the application of this method deals with road vehicles, a 3-D map of the drivable space has also been used to constrain the vehicle location to improve the performance of the method.

International audience; This paper describes a method which models the time correlation errors of a standalone L1-GPS receiver by integrating front-view camera measurements map-matched with a lane marking map. An identification method of the parameters of the shaping model is presented and evaluated with real data. The observability of the augmented state vector is demonstrated according to an algebraic definition. A positioning solver based on extended Kalman filtering with measured input is developed in the local ENU (East-North-Up) frame. This formulation allows including efficiently speed sensors in a dead-reckoned way. Outdoor experiments results carried out with a ground truth system are used to illustrate the performance of the proposed method. 1 INTRODUCTION Estimating the pose in real-time is a primary function for road automated vehicles. Recently, highly accurate digital maps have become key components for autonomous vehicle navigation and can enhance the positioning performance [1][2][3]. In this paper, we propose to use a lane marking map of the navigation area with centimetre-level precision. The map consists of two-lane roadways with dashed lane markings in the centre of the road and solid lane markings on the both sides of the road. Lane marking is expressed by polylines in a local frame near the navigation area. In order to exploit this lane marking map, we have integrated a camera from the Lane Departing Warning Systems of the experimental vehicle. The camera's lane marking detection module is able to measure the lateral distance between the detected lane marking and the vehicle in the camera frame. The camera can also identify the type of the detected lane markings (solid, dashed and road edge). Since the camera gives relative measurements, the next step is to relate the measures with the vehicle's state. For this, we have proposed a refined camera observation model which express camera measurement by a function of the state vector and the parameters of the detected lane marking [4]. So a data association process is necessary to associate a camera measurement with a certain lane marking segment in the map database. It is similar to a map matching process. The difference is that the map matching process can introduce a non-zero mean errors by simply projecting the vehicle position onto a segment which usually represents the road centreline. However, the camera has limitations. For example, there is no lane markings in some parts of the navigation area and the camera often fails to detect lane marking at crossroads. If the uncertainty domain grows too quickly during these times, it can be very challenging to distinguish different lanes when the camera retrieves measurements. Therefore Dead-Reckoning (DR) sensors are integrated to increase the performance particularly in terms of availability. We demonstrate the observability of the gyro bias thanks to the camera observation. It is important to note that the camera observation can only improve the positioning in lateral direction for a given moment, although the longitudinal accuracy is also improved if the road changes direction. GPS is very useful to compensate DR's long-term drift and it is mandatory for a cold start initialization of the localization system. However the L1-GPS positioning errors are not white and can be affected by strong biases, which are mainly due to atmosphere propagation delays, multi-path and real-time orbit errors, particularly in urban area. In this paper, we study GPS biases shaping models that are used in a filter by augmenting the state vector. The GPS errors are studied thanks to the use of a high-end localization system with centimetre-level accuracy. Autocorrelations of different sequences are analysed to find repeatable behaviour supposing that the bias is stationary (GPS measures with abrupt bias changing according to multi-path is supposed to be rejected by the localization solver). The bias is modeled by an autoregressive model. Parameters identification methods are discussed for integrating the shaping filter into the localization solver. The observability of GPS biases being not obvious, one of our contribution is the demonstration of the observability of the augmented state vector, which contains the pose of the vehicle, a gyro bias and GPS errors, according to an algebraic definition. The remainder is organized as follows. Section 2 introduces the GPS error modeling. Section 3 describes the system modeling and demonstrates the observability of the state vector. Outdoor experimental results are reported in Section 4. Section 5 concludes the paper.

7 pages; International audience; Intelligent vehicles refer currently to vehicles able to drive autonomously or able to provide pertinent information to the driver for safety, assistance and comfort. Cognitive cars are intelligent vehicles with additional capabilities like being able to collaborate with the driver in operating conditions. In this paper, a multi-agent system is used as a “digital butler” that does the interface between the driver and the machine. In order to test this approach, we consider an Advanced Driving Assistance System (ADAS) providing speed warnings when approaching dangerous areas. The system has been tested on an actual use case carried out with an experimental vehicle. We report some illustrative collaboration between the driver and the machine.

8 pages; International audience; This paper describes a new Map-Matching method relying on the use of Particle Filtering. Since this method implements a multi-hypothesis road-tracking strategy, it is able to handle ambiguous situations arising at junctions or when positioning accuracy is low. In this Bayesian framework, map-matching integrity can be monitored using normalized innovation residuals. An interesting characteristic of this method is its efficient implementation since particles are constraint to the road network; the complexity is reduced to one dimension. Experimental tests carried out with real data are finally reported to illustrate the performance of the method in comparison with a ground truth. The current real-time implementation allows map-matching at 100 Hz with confidence indicators which is relevant for many map-aided ADAS applications.

International audience; This paper proposes an evidential occupancy grid mapping framework that is designed to manage these different sources of uncertainty. An evidential sensor model interprets the data acquired by the Velodyne and maps it to a 2D evidential map. The information from the sensor is processed based on the least commitment principle to guarantee information integrity. Experimental results prove that this approach can handle efficiently the uncertainties of the sensor and thus a highly reliable ground truth map all around the vehicle can be built.

Evidential grids have recently shown interesting properties for mobile object perception. Evidential grids are a generalisation of Bayesian occupancy grids using Dempster- Shafer theory. In particular, these grids can handle efficiently partial information. The novelty of this article is to propose a perception scheme enhanced by geo-referenced maps used as an additional source of information, which is fused with a sensor grid. The paper presents the key stages of such a data fusion process. An adaptation of conjunctive combination rule is presented to refine the analysis of the conflicting information. The method uses temporal accumulation to make the distinction between stationary and mobile objects, and applies contextual discounting for modelling information obsolescence. As a result, the method is able to better characterise the occupied cells by differentiating, for instance, moving objects, parked cars, urban infrastructure and buildings. Experiments carried out on real- world data illustrate the benefits of such an approach., Comment: 6 pp. arXiv admin note: substantial text overlap with arXiv:1207.1016

Le manuscrit est organisé autour de 4 chapitres scientifiques et techniques. Le premier porte sur la radiolocalisation. Il en dresse un état de l'art en décrivant les mesures et le fonctionnement de systèmes de radiolocalisation sur la base des exemples du LORAN C et du GPS. Ces principes serviront aux modélisations du chapitre 4. Les amers utilisés par les systèmes de localisation ont été souvent peu considérés. Depuis quelques années et suite au progrès des techniques de cartographie et localisation simultanées, ils sont au centre des préoccupations de chercheurs. Ainsi dans le chapitre 3, je présente quelques types d'amers en cherchant à faire le lien avec l'information géographique et les bases de données navigables utiles pour leur gestion. Le chapitre 4 pose le problème de la localisation dynamique comme un problème d'observation d'état et décrit les différentes méthodologies que j'ai étudiées.C'est le chapitre le plus méthodologique dans lequel je présente le plus de résultats. Enfin, le chapitre 5 présente l'application de mes travaux à l'aide à la conduite automobile.Le document se terminera par une synthèse de mes contributions scientifiques et dressera des perspectives de recherche.

International audience; Multipath signals have a strong effect on position estimates; these are studied intensely theoretically and experimentally. In this paper a domain oriented perspective is presented, passenger vehicles traversing dense urban areas. The purpose is to examine the effects of multipath on pseudo-ranges and Doppler measurements. The approach combines a theoretical formulation with extensive testing. The experiment includes the acquisition of measurements provided by a low-cost automotive GPS receiver on-board a vehicle in real traffic conditions. In addition, the vehicle trajectory was measured precisely to allow for the quantification of measurement errors. Two approaches are used in the experimental part to determine the pseudo-range and Doppler errors: Vehicle on-board measurements plus post-processing; vehicle onboard measurements, data from an external GPS reference station and post-processing. The results enable the identification of satellites whose signals are corrupted by multipath, to plot the results with respect to the trajectory so as to provide contextual information that facilitates understanding. It was also possible to establish the discontinuous relationship between signal to noise ratio and pseudo-range & Doppler errors.

International audience; Integrity is an attribute of the quality of performance of a positioning technology that is getting crucial for many Intelligent Transportation Systems dealing with safety. Managing the confidence in estimated locations is difficult when operating in complex outdoors environments, particularly for Global Navigation Satellite Systems (GNSS). A common integrity process consists in removing outliers amongst the measurements using Fault Detection and Exclusion (FDE) algorithms and then to compute Protection Levels (PL) that quantify the integrity zone. This snapshot paradigm is applicable mainly to the aerospace domain where high pseudo-ranges redundancy exists, which is not the case for land applications. In this paper, a new approach to localization integrity for land mobile robot is proposed by combining vehicle and GNSS data, stored within a short-term memory, called data horizon. This formulation allows for the application of FDE algorithms on the combined data set. It also leads to a gain of redundancy that enables PLs to be reduced and to increase the availability of integrity algorithms. Results from evaluation tests are included to demonstrate the validity of the approach.

International audience; This paper studies the benefits of integrating proprioceptive measurements and a-priory geographical knowledge in the GNSS computation to increase availability and accuracy for modern ITS applications. A tightly-coupled framework for merging GNSS pseudo-ranges and odometric measurements is first introduced. An efficient way to merge a 2D navigable road-map is then presented, including the road selection and the map measurement model. A cautious fusion strategy, which provides enhanced GNSS-like positioning is performed, using an Extended Kalman filter monitoring innovation signals. This method has been applied to real-field data recorded during spring 2007. Results show the benefits of the proposed framework for GPS-computation under adverse conditions, i.e. with bad satellites availability and configuration.

International audience; In this paper, we propose a tightly-coupled map-matching algorithm without explicit road selection stage. By combining a geometrical and a topological approach, multi-hypothesis road tracking is achieved using a constraint Rao-Blackwellised particle filter that tightly merges raw GNSS measurements --Pseudo-ranges and Dopplers-- with a 3D navigable road-map.

International audience; Our approach, using a set-inversion method in a bounded-error context, enables to compute a position fix in the form of a localization area taking into account all sources of error (notably space vehicles position errors) without linearizing. The method computes an integrity zone recursively using bisection. Robustness can also be achieved with several simultaneous faulty measurements. A way to set bounds on measurements to meet an integrity risk requirement is presented. An experimental validation using real L1 code measurements is also presented in the paper.

International audience; Efficient and reliable map matching algorithms are essential for Advanced Driver Assistance Systems. While most of the existing solutions fail to provide trustworthy outputs when the situation is ambiguous (road intersections, roundabouts, parallel roads ...), we present in this paper a new map-matching method based on a multi-hypothesis road tracking that takes advantage of the geographical database road connectedness to provide a reliable road-matching solution with a confidence indicator that can be used for integrity monitoring purposes

A very important parameter to measure the quality of localisation systems is integrity in Global Navigation Satellite System (GNSS). Within this perspective, Integrity first considers the removal of outliers at the receiver level. It is typically monitored applying Fault Detection and Exclusion (FDE) algorithms. Then, integrity ensures a high probability bounded position error by computing protection levels. However, their performance depends very much on the faults characteristics and data redundancy which reflect the environment in which location estimates are made, and thus differ from scenario to scenario. After analyzing the specificities of GNSS based positioning in the automotive context, and the mechanism of the Receiver Autonomous Integrity Monitoring (RAIM) algorithm, a novel Kalman Filter based FDE algorithm is proposed in this paper. This is validated using hybrid simulation techniques based on real data from vehicle proprioceptive sensors and partially simulated GNSS data, where satellite configuration and faults characteristics can be controlled. The chosen scenario is that found in urban environments, representative of areas where localisation systems suffer many difficulties when mounted on land platforms. A comparison between the RAIM and the new algorithm is included.

International audience; Global Navigation Satellite Systems (GNSS) are often used to localize a receiver with respect to a given map. This association problem, also known as map-matching, is usually addressed using estimated positions computed by the GNSS receiver. In this paper we propose a method that combines the cartographic data in the GNSS computation fix itself. We focus on the use of a road network provided by cartographers such as NavTeQ or TeleAtlas. Geo-referenced data is modelled by segments that can be used as constraints or fused with the pseudo-ranges. Using residuals, a new method for tackling the underlying problem of the road selection is proposed. We show that this approach is also well adapted to the integrity problem of map-matching, since a consistency test is derived. Experimental results illustrate the performance of this method with different maps.

International audience; In this paper, we present an extensive study about the influence and the behavior of the intrinsic camera calibration imprecisions and their propagation into the extrinsic calibration between a camera and a multi-layer lidar. Usually, the extrinsic calibration process involves the pose estimation of a reference object in the Cartesian frame of each sensor. From this fact, it is necessary to know the camera intrinsic parameters for estimating the position of an object on an image. Therefore, the extrinsic calibration process is directly dependent of the intrinsic camera parameters. By using different projective camera models, we estimate the influence with respect to the computation of the extrinsic parameters.

This paper describes the system approach for a future positioning system to be developed for the European CVIS project (www.cvisproject.org) that deals with an increased transport safety and efficiency using advanced positioning, mapping and communication technologies. Positioning and mapping functions will be provided by the sub-project POMA, where new methods for the combination of several positioning components will be developed. The whole positioning system setup of CVIS will then be deployed at DLR in Berlin. This setup will not only cover onboard components but also infrastructure based sensors and sensor networks together with enhanced positioning algorithms. This new approach shall enable high accuracy and reliability for the following field trials spread over Europe.

International audience; This paper presents a road selection strategy for novel road-matching methods that are designed to support real-time navigational features within Advanced Driving-Assistance Systems (ADAS). Selecting the most likely segment(s) is a crucial issue for the road-matching problem. The selection strategy merges several criteria using Belief theory. Particular attention is given to the development of belief functions from measurements and estimations of relative distances, headings, and velocities. Experimental results using data from antilock brake system sensors, the differential Global Positioning System receiver, and the accurate digital roadmap illustrate the performances of this approach, particularly in ambiguous situations.

This paper tackles the problem of precise car localization for Advance Driver Assistance System applications. Localization is performed by loosely coupling proprioceptive data with exteroceptive perceptions. Two localization systems were designed, one based on Differential GPS and the other on Real Time Kinematic GPS. In both cases, sensor latencies were precisely measured and taken into account to minimize their effects. Track experimentations were carried out with two vehicles at different speeds. A methodoly based on a comparison with a Post-Processed Kinematic GPS reference is presented to evaluate performances of the two systems.

International audience; This paper describes a novel road-matching method designed to support the real-time navigational function of cars for advanced systems applications in the area of driving assistance. This method provides an accurate estimation of position for a vehicle relative to a digital road map using Belief Theory and Kalman filtering. Firstly, an Extended Kalman Filter combines the DGPS and ABS sensor measurements to produce an approximation of the vehicle's pose, which is then used to select the most likely segment from the database. The selection strategy merges several criteria based on distance, direction and velocity measurements using Belief Theory. A new observation is then built using the selected segment, and the approximate pose adjusted in a second Kalman filter estimation stage. The particular attention given to the modeling of the system showed that incrementing the state by the bias (also called absolute error) of the map significantly increases the performance of the method. Real experimental results show that this approach, if correctly initialized, is able to work over a substantial period without GPS.

This thesis is about the design of driving assistance systems for level 3 urban automated driving. Because of a more complex of the environment and a larger set of possible trajectories, the algorithms of highway automated driving are not adapted to urban environment. This thesis objective is to provide methods and algorithms to enable the vehicle to perform automated driving in urban scenarios, focusing on the vehicle lateral guidance and on the path planning. The proposed solutions operate in real-time on board of the automated vehicle prototypes. The contribution of this thesis is as theoretical as practical.After a synthesis of the driving assistance systems available on current cars and a presentation of the prototypes used for the validation of the algorithms developed in the thesis, a complete analysis of the vehicle lateral dynamics is carried out in time and frequency domains. This analysis enables the design of observers of the lateral dynamics in order not only to estimate signals required for the lateral guidance functions but also to increase reliability of available measurements. Based on the conclusions from the analysis of lateral dynamic, a multi-controller solution has been proposed. It enables the computation of a steering wheel input to follow a trajectory at any longitudinal speed. The solution is validated in simulation and on real road traffic for lane change scenarios. Another contribution consist in an analysis on the other vehicles of the environment is conducted in order to identify their behaviors and which maneuver there are performing. This analysis is essential for the path planning function developed in the thesis. This method, based on the A* algorithm and extended to respect geometric and dynamic constraints, firstly focuses on static environment such as a parking lot. Waypoints are added to the method in order to compute trajectories compatible with traffic regulation and improve the computation time. The method is then adapted for dynamic environment where, in the end, the vehicle is able to perform overtaking manoeuvers in a complex environment., Cette thèse s’inscrit dans le développement et la conception de fonctions d’aide à la conduite pour les véhicules autonomes de niveau 3 et plus en milieu urbain ou péri urbain. Du fait d’un environnement plus complexe et de trajectoires possibles plus nombreuses et sinueuses, les algorithmes des véhicules autonomes développés pour l’autoroute ne sont pas adaptés pour le milieu urbain. L’objectif de la thèse est de mettre à disposition des méthodes et des réalisations pour permettre au véhicule autonome d’évoluer en milieu urbain. Cette thèse se focalise sur la proposition de solutions pour améliorer le guidage latéral des véhicules autonomes en milieu urbain à travers l’étude de la planification de trajectoire en situation complexe, l’analyse du comportement des usagers et l’amélioration du suivi de ces trajectoires complexes à faibles vitesses. Les solutions proposées doivent fonctionner en temps réel dans les calculateurs des prototypes pour pouvoir ensuite être appliquées sur route ouverte. L’apport de cette thèse est donc autant théorique que pratique.Après une synthèse des fonctions d’aide à la conduite présentes à bord des véhicules et une présentation des moyens d’essais mis à disposition pour la validation des algorithmes proposés, une analyse complète de la dynamique latérale est effectuée dans les domaines temporel et fréquentiel. Cette analyse permet alors la mise en place d’observateurs de la dynamique latérale pour estimer des signaux nécessaires aux fonctions de guidage latéral et dont les grandeurs ne sont pas toujours mesurables, fortement dégradées ou bruitées. La régulation latérale du véhicule autonome se base sur les conclusions apportées par l’analyse de cette dynamique pour proposer une solution de type multirégulateur capable de générer une consigne en angle volant pour suivre une trajectoire latérale quelle que soit la vitesse. La solution est validée tant en simulation que sur prototype pour plusieurs vitesses sur des trajectoires de changement de voie. La suite de la thèse s’intéresse à la génération d’une trajectoire en milieu urbain tenant compte non seulement de l’infrastructure complexe (intersection/rond-point) mais également des comportements des véhicules autour. C’est pourquoi, une analyse des véhicules de l’environnement est menée afin de déterminer leur comportement et leur trajectoire. Cette analyse est essentielle pour la méthode de génération de trajectoire développée dans cette thèse. Cette méthode, basée sur l’algorithme A* et enrichie pour respecter les contraintes géométriques et dynamiques du véhicule, se focalise d’abord dans un environnement statique complexe de type parking ou rond-point. Des points de passage sont intégrés à la méthode afin de générer des trajectoires conformes au code de la route et d’améliorer le temps de calcul. La méthode est ensuite adaptée pour un environnement dynamique où le véhicule est alors capable, sur une route à double sens de circulation, de dépasser un véhicule avec un véhicule arrivant en sens inverse.

Swarm behavior refers to individuals travelling in a group and using only localknowledge of their environment.The scientific objective of the thesis is to implement this type of behaviormodel to vehicles traveling on road, in order to assist the driver in his actions for bothits comfort and security.From a literature review, a prioritization strategy was set up to create anAdvanced Driver Assistance System (ADAS). At first, it is to generate a path from thistype of model that respects the motorway constraints. Then, vehicle dynamics istaken into account in order to transmit to the driver through cruise control and hapticfeedback steering wheel, both based on the CRONE control, maneuvers needed tofollow this trajectory. Finally, the driver assistance system is not only implemented ona dynamic driving simulator to gather driver’s feelings but it is also implemented intraffic simulation software to evaluate gains obtained for a set of equipped vehicles., Le mouvement en essaim est défini par l'action d'un ensemble d'individusautopropulsés se déplaçant en groupe uniquement à l’aide de la connaissance locale de leur environnement.L'objectif scientifique de la thèse consiste à mettre en oeuvre ce type demodèle de comportement appliqué à un flot de véhicules se déplaçant sur un profilroutier, et ce afin d'assister le conducteur dans ses actions à la fois pour son confortet sa sécurité.A partir de l’analyse d’une synthèse bibliographique, une stratégie dehiérarchisation a été mise en place afin de créer un système d’aide à la conduite ouADAS (de l’anglais « Advanced Driver Assistance System »). Ainsi, dans un premiertemps, il s’agit de générer une trajectoire à partir de ce type de modèle qui respecteles contraintes autoroutières. Ensuite, la dynamique du véhicule est prise en compteafin de transmettre au conducteur via une régulation de vitesse et un retour haptiqueau volant, les deux étant basés notamment sur la commande CRONE, lesmanoeuvres nécessaires au suivi de cette trajectoire. Enfin, le système d’aide à laconduite est mis en oeuvre, non seulement sur un simulateur dynamique de conduiteafin de recueillir le ressenti des conducteur, mais aussi au sein d’un logiciel desimulation de trafic pour évaluer les gains obtenus dans le cas d’un ensemble devéhicules équipés.

International audience; Being able to predict the trajectory of a human driver's potential lane change behavior in urban high way scenario is crucial for lane change risk assessment task. A good prediction of the driver's lane change trajectory makes it possible to evaluate the risk and warn the driver beforehand. Rather than generating such a trajectory only using a mathematical model, this paper develops a lane change trajectory prediction approach based on real human driving data stored in a database. In real-time, the system generates parametric trajectories by interpolating k human lane change trajectory instances from the pre-collected database that are similar to the current driving situation. In order to build this real lane change database, a human lane change data collection vehicle platform is developed. Extensive experiments have been carried out in urban highway environments to build a significant database with more than 200 lane changes. Real results show that this approach produces lane change trajectories that are quite similar to real ones which makes it suitable to predict humanlike lane change maneuvers.