abstract: Laser radars or lidar’s have been used extensively to remotely study winds within the atmospheric boundary layer and atmospheric transport. Lidar sensors have become an important tool within the meteorology and the wind energy community. For example, Doppler lidars are used frequently in wind resource assessment, wind turbine control as well as in atmospheric science research. A Time of Flight based (ToF) direct detection lidar sensor is used in vehicles to navigate through complex and dynamic environments autonomously. These optical sensors are used to map the environment around the car accurately for perception and localization tasks that help achieve complete autonomy. This thesis begins with a detailed discussion on the fundamentals of a Doppler lidar system. The laser signal flow path to and from the target, the optics of the system and the core signal processing algorithms used to extract velocity information, were studied to get closer to the hardware of a Doppler lidar sensor. A Doppler lidar simulator was built to study the existing signal processing algorithms to detect and estimate doppler frequency, and radial velocity information. Understanding the sensor and its processing at the hardware level is necessary to develop new algorithms to detect and track specific flow structures in the atmosphere. For example, the aircraft vortices have been a topic of extensive research and doppler lidars have proved to be a valuable sensor to detect and track these coherent flow structures. Using the lidar simulator a physics based doppler lidar vortex algorithm is tested on simulated data to track a pair of counter rotating aircraft vortices. At a system level the major components of a time of flight lidar is very similar to a Doppler lidar. The fundamental physics of operation is however different. While doppler lidars are used for radial velocity measurement, ToF sensors as the name suggests provides precise depth measurements by measuring time of flight between the transmitted and the received pulses. The second part of this dissertation begins to explore the details of ToF lidar system. A system level design, to build a ToF direct detection lidar system is presented. Different lidar sensor modalities that are currently used with sensors in the market today for automotive applications were evaluated and a 2D MEMS based scanning lidar system was designed using off-the shelf components. Finally, a range of experiments and tests were completed to evaluate the performance of each sub-component of the lidar sensor prototype. A major portion of the testing was done to align the optics of the system and to ensure maximum field of view overlap for the bi-static laser sensor. As a laser range finder, the system demonstrated capabilities to detect hard targets as far as 32 meters. Time to digital converter (TDC) and an analog to digital converter (ADC) was used for providing accurate timing solutions for the lidar prototype. A Matlab lidar model was built and used to perform trade-off studies that helped choosing components to suit the sensor design specifications. The size, weight and cost of these lidar sensors are still very high and thus making it harder for automotive manufacturers to integrate these sensors into their vehicles. Ongoing research in this field is determined to find a solution that guarantees very high performance in real time and lower its cost over the next decade as components get cheaper and can be seamlessly integrated with cars to improve on-road safety. / Dissertation/Thesis / Doctoral Dissertation Mechanical Engineering 2018

L'augmentation continue du nombre de véhicules a sans cesse conduit à innover pour mieux réguler le trafic routier. Aujourd'hui, l'automatisation des véhicules et la communication inter-véhiculaire rend possible une nouvelle approche : l'intersection autonome et coopérative (IAC) dans laquelle les véhicules autonomes communiquent pour négocier leur droit de passage à l'intersection. Théoriquement plus efficace, plus sûre et plus économique que les approches traditionnelles, son implémentation et son déploiement constituent néanmoins un défi scientifique et technologique. Dans ce mémoire, nous proposons plusieurs contributions soutenant ce développement : une stratégie de contrôle longitudinal et latéral du véhicule conçues pour fonctionner avec des informations discrètes et présentant une latence (provenant de la communication sans-fil ou du positionnement satellite), une solution de prévention des risques d'inter-blocage spécifique à l'IAC, et enfin, l'implémentation et l'expérimentation d'un prototype réel d'IAC avec trois véhicules autonomes. Un travail exploratoire est ensuite fourni pour généraliser les principes de l'IAC, afin d'automatiser et optimiser les opérations de changement de voie et d'insertion. / The continuous increase in the number of vehicles has constantly led to innovation to improve the regulation of road traffic. Today, vehicle automation and inter-vehicular communication make possible a new approach: the autonomous and cooperative intersection (ACI) in which autonomous vehicles communicate to negotiate their right of way at the intersection. Theoretically more effective, safer and more economical than traditional approaches, its implementation and deployment constitute a scientific and technological challenge. In this paper, we propose several contributions supporting this development: a longitudinal and lateral vehicle control strategy designed to work with information discrete and suffering from latency (coming from wireless communication or satellite positioning), a preventive solution of the risks of deadlock specific to the ACI, and finally the implementation and testing of a real prototype of ACI with three autonomous vehicles. An exploratory work is then provided to generalize the principles of the ACI, in order to automate and optimize the operations of lane change and insertion.

Dans ce travail, nous exploitons le potentiel offert par les véhicules autonomes coopératifs, pour fluidifier le trafic dans une intersection isolée puis dans un réseau d’intersections. Nous proposons le protocole SVAC (Système du Véhicule Actionneur Coopératif) permettant de réguler une intersection isolée. SVAC est basé sur une distribution individuelle du droit de passage qui respecte un ordre précis donné par une séquence de passage.Pour optimiser la séquence de passage, nous définissons la politique PED (Politique d’Evacuation Distribuée) permettant d’améliorer le temps d’évacuation total de l’intersection. La création de la séquence de passage est étudiée à travers deux modélisations. Une modélisation sous forme de graphes permettant le calcul de la solution optimale en connaissant les dates d’arrivée de tous les véhicules, et une modélisation de type réseaux de Petri avec dateurs pour traiter la régulation temps-réel. Des tests réels avec des véhicules équipés ont été réalisés pour étudier la faisabilité du protocole SVAC. Des simulations mettant en jeu un trafic réaliste permettent ensuite de montrer la capacité de fluidifier le trafic par rapport à une régulation classique par feux tricolores.La régulation d’un réseau d’intersections soulève les problèmes d’interblocage et de réorientation du trafic. Nous proposons le protocole SVACRI (Système du Véhicule Actionneur Coopératif pour les Réseaux d’Intersections) qui permet d’éliminer à priori les risques d’interblocage à travers la définition de contraintes d’occupation et de réservation de l’espace et du temps. Nous étudions également la possibilité d’améliorer la fluidité du trafic à travers le routage des véhicules, en tirant avantage du protocole SVACRI. Enfin, nous généralisons le système de régulation proposé pour la synchronisation des vitesses aux intersections. / The objective of this work is to use the potential offered by the wireless communication and autonomous vehicles to improve traffic flow in an isolated intersection and in a network of intersections. We define a protocol, called CVAS (Cooperative Vehicle Actuator System) for managing an isolated intersection. CVAS distributes separately the right of way to each vehicle according to a specific order determined by a computed sequence.In order to optimize the sequence, we define a DCP (Distributed Clearing Policy) to improve the total evacuation time of the intersection. The control strategy is investigated through two modeling approaches. First graph theory is used for calculating the optimal solution according to the arrival times of all vehicles, and then a timed Petri Net model is used to propose a real-time control algorithm. Tests with real vehicles are realized to study the feasibility of CVAS. Simulations of realistic traffic flows are performed to assess our algorithm and to compare it versus conventional traffic lights.Managing a network of intersections raises the issue of gridlock. We propose CVAS-NI protocol (Cooperative Vehicle actuator system for Networks of Intersections), which is an extension of CVAS protocol. This protocol prevents the deadlock in the network through occupancy and reservation constraints. With a deadlock free network we extend the study to the traffic routing policy. Finally, we generalize the proposed control system for synchronizing the vehicle velocities at intersections.

Le travail de thèse de Baudoin Dafflon se situe dans la thématique générale des systèmes auto-organisés, dans la version centrée sur les systèmes multi-agents. Ce travail repose sur l’idée que l’organisation d’un système distribué peut émerger à partir du comportement de chacune des d’entités autonomes et proactives qui le composent. L’organisation émergente satisfait l’objectif confié au système auto-organisé. Cette thèse s’est particulièrement occupé de domaines applicatifs où les systèmes auto-organisés sont formés par des entités matérielles intelligents, en général mobiles, situées dans un environnement physique relativement incertain. Ses études de cas appartiennent à la problématique de la conduite en convoi de flottes de véhicules autonomes.En particulier, B. Dafflon s’est intéressé à la prise de décision pour ce genre de systèmes. Cette question est particulièrement importante lorsque l’environnement est incertain, soit par ce que les entités autonomes en ont une connaissance imparfaite, soit par ce que des phénomènes aléatoires peuvent se produire. Dans ce cas chaque entité doit décider, de façon décentralisée quelle est la meilleure action locale à exécuter pour que l’organisation globale soit préservée. Si bien l’échange d’informations entre entités n’est pas exclu d’emblée pour ce type de systèmes, les travaux de cette Thèse ont été conduits en adoptant la contrainte de non-communication, dans le but d’explorer la portée des approches purement autonomes et locales de la prise de décision par les agents.Parmi les contributions substantielles apportées par cette thèse nous pouvons mentionner les suivants. Premièrement, la proposition d’une architecture qui sépare la fonction décision de la fonction contrôle, ce qui permet de mieux intégrer des fonctionnalités qui prennent en compte l’incertain. Un exemple dans le domaine applicatif (la conduite en convoi) serait l’évitement d’obstacles. Deuxièmement, B. Dafflon propose l’introduction d’une hiérarchie de systèmes multi-agents où la prise de décision est également envisagée comme un processus auto-organisé dans un monde virtuel d’agents. Ce monde virtuel est défini par un processus d’abstraction des informations données par les percepts des agents matériels. Finalement, cette Thèse propose des modèles d’interaction pour ces agents virtuels qui reposent sur des processus physiques (lois d’attraction-répulsion, dynamique des gazes, ...). L’intérêt en est que la décision peut être élaborée de façon calculatoire, en évaluant des indicateurs globaux inspirés des processus physiques modélisés.Les travaux présentés dans cette thèse ont été effectués et ont été financés par le projet ANR-VTT SafePlatoon. / Baudoin Dafflon’s doctoral work is placed in the general frame of self-organised systems. This kind of systems bases on the hypothesis that their organizationcan be the result of the individual behaviour of each one of the autonomous and proactive component entities. This emergent organization satisfies the objectives entrusted to the self-organizing system. This thesis is mainly directed to application domains in which the self-organizing system is composed of intelligent, frequently mobile material devices places in a physical, uncertain environment. Case-studies made included in this Thesis belong to the vehicle platoon problem.This Thesis interests mainlyto the decision-making problem of the self organizing systems of this kind. This aspect is particularly important when the system evolves in an uncertain environment. Uncertainty can be the consequence of an imperfect knowledge of the environment or because of the occurrence of unpredictable events. In those situations the autonomous entities have to decide locally about the best action to be performed in order for the global organization to be maintained. Even if information exchange among autonomous entities is not excluded a-priori, this doctoral work has been done under the restriction of non-communication, in order to investigate the reach of completely de-centralised approaches.Among the contributions of this thesis we can mention the following. Firstly, the proposal of an architecture that separates decision making and control, thereby allowing a simpler integration of uncertainty-management. An example from the application domain (vehicle platoons) is obstacle avoidance. Secondly, this thesis introduces a hierarchy of multi-agent systems such that decision-making is also conceived as a self-organized process in a virtual agent’s world. The later is built from an abstraction process based on the perceptions produced by material agents. Lastly, this thesis proposes physics-inspired interaction mechanisms for virtual agents. Those interaction models based on phenomena such as attraction-repulsion forces or gas dynamics, allow for a computational decision making process based on the laws and properties of the adopted physical models.Doctoral work presented in this Thesis has been done in the framework of the ANR-VTT project.

Soft computing techniques are generally well-suited for vehicular control systems that are usually modeled by highly nonlinear differential equations and working in unstructured environment. To demonstrate their applicability, two intelligent controllers based upon fuzzy logic theories and neural network paradigms are designed for performing a wall-following task and an autonomous parking task. Based on performance and flexibility considerations, the two controllers are implemented onto a reconfigurable hardware platform, namely a Field Programmable Gate Array (FPGA). As the number of comparative studies of these two embedded controllers designed for the same application is limited in the literature, one of the main goals of this research work has been to evaluate and compare the two controllers in terms of hardware resource requirements, operational speeds and trajectory tracking errors in following different pre-defined trajectories. The main advantages and disadvantages of each of the controllers are presented and discussed in details. Challenging issues for implementation of the controllers on the FPGA platform are also highlighted. As the two controllers exhibit benefits and drawbacks under different circumstances, this research suggests as well a hybrid controller scheme as an attempt to integrate the benefits of both control units. To evaluate its performance, the hybrid controller is tested on the same pre-defined trajectories and the corresponding results are compared to that of the fuzzy logic and the neural network based controllers. For further demonstration of the capabilities of the wall-following controllers in other applications, the fuzzy logic and the neural network controllers are used in a parallel parking system. We see this work to be a stepping stone for further research work aiming at real world implementation of the controllers on Application Specified Integrated Circuit (ASIC) type of environment.

Soft computing techniques are generally well-suited for vehicular control systems that are usually modeled by highly nonlinear differential equations and working in unstructured environment. To demonstrate their applicability, two intelligent controllers based upon fuzzy logic theories and neural network paradigms are designed for performing a wall-following task and an autonomous parking task. Based on performance and flexibility considerations, the two controllers are implemented onto a reconfigurable hardware platform, namely a Field Programmable Gate Array (FPGA). As the number of comparative studies of these two embedded controllers designed for the same application is limited in the literature, one of the main goals of this research work has been to evaluate and compare the two controllers in terms of hardware resource requirements, operational speeds and trajectory tracking errors in following different pre-defined trajectories. The main advantages and disadvantages of each of the controllers are presented and discussed in details. Challenging issues for implementation of the controllers on the FPGA platform are also highlighted. As the two controllers exhibit benefits and drawbacks under different circumstances, this research suggests as well a hybrid controller scheme as an attempt to integrate the benefits of both control units. To evaluate its performance, the hybrid controller is tested on the same pre-defined trajectories and the corresponding results are compared to that of the fuzzy logic and the neural network based controllers. For further demonstration of the capabilities of the wall-following controllers in other applications, the fuzzy logic and the neural network controllers are used in a parallel parking system. We see this work to be a stepping stone for further research work aiming at real world implementation of the controllers on Application Specified Integrated Circuit (ASIC) type of environment.

In this research, an autonomous vehicle function management methodology is studied. In accordance with the traffic situation, the decision making level chooses the optimal function that guarantees safety and minimizes fuel consumption while the control level is implemented via neuromorphic strategy based on the brain limbic system. To realize the decision making strategy, the Analytic Hierarchy Process (AHP) is used by considering driving safety, driving speed, and fuel efficiency as the objectives. According to the traffic situation and predefined driving mode, Lane Change Maneuver (LCM) and Adaptive Cruise Control (ACC) are chosen as the alternative functions in the AHP framework. The adaptive AHP is utilized to cope with dynamically changing traffic environment. The proposed adaptive AHP algorithm provides an optimal relative importance matrix that is essential to make decisions under a varying traffic situation and driving modes. The simulation results show that proposed autonomous vehicle function management structure produces optimal decisions that satisfy the driving preference. The stability of BLS based control is also investigated via Cell-to-Cell Mapping. In this research, autonomous vehicle functions such as Lane change maneuver and Adaptive cruise control are developed by means of BLS based control. The simulation results considered various traffic situations that an autonomous vehicle can encounter. To demonstrate the suggested control method Cell-to-Cell Mapping is utilized. Subsequently, the autonomous vehicle function management strategy is developed by Applying AHP and an adaptive AHP strategy is developed to cope with various traffic situations and driving modes. The suggested method is verified numerical simulations.

In this thesis, video frames acquired by a camera in a moving car are processed for detection of candidates of triangular, rectangular and circular traffic/road signs based on mainly shape information by performing contour analysis. Color information is utilized as an auxiliary method to improve detection. Then recognition based on template matching is realized on detected traffic/road sign candidates. Detection and recognition results of traffic/road signs in video frames taken in different time intervals of day for these methods are compared. After implementation, results show that the video scene taken in a sunny day in the afternoon gives better results than others. Binary threshold plays a great role in detection with respect to Canny edge detector especially for triangular and rectangular traffic signs. Higher number of binary threshold levels improves detection in general. In addition, the recognition rate for triangular and rectangular traffic/road signs is higher than that of circular sings in general by the methods used in this study.

This dissertation solves variations of three mathematical problems for autonomous vehicle formations: (1) formation shape control in the plane, (2) robust information architecture design, and (3) formation attitude synchronization. An autonomous vehicle formation is a collection of vehicles, each with computation, communication, sensing, and control capabilities, that cooperate to achieve a common objective. Accelerating advancements are making possible a range of science and engineering applications, such as satellite formations for deep-space imaging, teams of unmanned aircraft for military reconnaissance and surveillance missions, and submarine swarms for oceanic exploration. The ubiquitous potential of these applications is driving theoretical work on autonomous vehicle formations across a range of disciplines. A major theoretical question in the field of control theory, and the main focus of this dissertation, is how the properties of the information architecture (i.e. a mapping of the information flow amongst the agents), relate to the stability properties of the desired shape and orientation under certain control laws. A secondary focus is how to design the information flow so that loss of an agent does not destroy the formation's ability to maintain a desired shape. As a motivating example, a solution to a coordinated standoff tracking problem is presented to demonstrate how an instance of a class of information architectures, which are called persistent and related to rigid graph theory, can be used to achieve a formation objective in a practical scenario involving a team of unmanned aircraft. A generalized formation shape control problem is then solved for a class of persistent architectures. This solution gives only local stability results; global stability is analyzed for a four-agent formation and several open problems are identified. The problem of agent loss is addressed by performing a self-repair operation in the event of agent loss and separately by designing robustness into the information architecture a priori. Finally, a rigid body attitude synchronization problem with communication time delays is solved for a class of information architectures based on spectral graph theory. / text

Through the work of researchers and the development of commercially availableproducts, automated guidance has become a viable option for agricultural producers.Some of the limitations of commercially available technologies are that they onlyautomate one function of the agricultural vehicle and that the systems are proprietary toa single machine model.The objective of this project was to evaluate a controller area network (CAN bus)as the basis of an automated guidance system. The prototype system utilized severalmicrocontroller-driven nodes to act as control points along a system wide CAN bus.Messages were transferred to the steering, transmission, and hitch control nodes from atask computer. The task computer utilized global positioning system data to determinethe appropriate control commands.Infield testing demonstrated that each of the control nodes could be controlledsimultaneously over the CAN bus. Results showed that the task computer adequatelyapplied a feedback control model to the system and achieved guidance accuracy levelswell within the range sought. Testing also demonstrated the system's ability tocomplete normal field operations such as headland turning and implement control.