Ce travail porte sur la conception d’un système capable d’effectuer des manœuvres de parking automatique plus polyvalent que ceux actuellement commercialisés, tout en conservant une définition technique des capteurs extéroceptifs limités en prix et en gabarit. Un cas d’usage typique est de permettre au véhicule de se rendre automatiquement dans la zone de garage du domicile de son propriétaire, cette fonction est classiquement appelée voiturier autonome à domicile. Partant de l’existant et connaissant les performances attendues, une architecture système et une architecture fonctionnelle ont été tracées. Cela a permis de constituer un ensemble de fonctions interconnectées qui ont participé dans la création d’une architecture software modulaire ainsi que dans la création des interfaces de connexion au véhicule prototype. Dans un premier temps, nous explorons la problématique de la détection d’obstacles. Partant d’un système propriétaire fermé de capteurs ultrason, nous avons réussi à réaliser une carte d’obstacle à un niveau de précision supérieur au produit d’origine. Une augmentation de la limite de détection des capteurs ultrason a été réalisée utilisant une technique Structure from Motion. Ces informations d’occupation ont été exploitées par la suite pour traiter la problématique de détection du couloir de navigation. Dans un second temps, la fonction de localisation du véhicule est abordée. Trois techniques de localisation collaborent pour une robustesse de fonctionnement continu : la localisation odométrique, la localisation par appariement des grilles d’occupation et la localisation par appariement entre une image actuelle et une base d’images adaptée à notre besoin et améliorée en termes de temps de calcul. Enfin, nous nous sommes intéressés à la problématique de navigation du véhicule. Nous avons considéré résolue la problématique de contrôle des actionneurs pour le suivi d’une trajectoire donnée et nous nous sommes concentrés sur la création d’une trajectoire admissible. Nous avons développé une technique de planification locale pour l’évitement d’un d’obstacles non cartographiés. Pour la construction de trajectoire nous avons utilisé des courbes à géométrie connue et avons montré qu’en utilisant trois clothoïdes et éventuellement deux arcs de cercle (si le braquage maximal est atteint) il est possible de créer des trajectoires à courbure continue adaptées à notre situation. Nous avons montré que l’utilisation d’une carte d’obstacles nous permet de prédire plus en avance de la possibilité d’emprunter un certain couloir de navigation. Chacune des parties de ce travail a fait l’objet de validations en simulation mais aussi sur des données réelles démontrant la pertinence des approches proposées quant à l’application visée. / This work covers the conception of a system capable to do automatic parking maneuvers more versatile than those already commercialized, respecting the technical definition of exteroceptive sensors limited by costs and weight. A typical use case is to set a vehicle to park autonomously in the parking lot of a home, function generally called autonomous home valet parking. Taking from the existing and knowing the expected performances, a system architecture and a functional architecture were drawn. This allowed to compose an assembly of interconnected functions that participated in the creation of modular software architecture, as well as in the creation of connection interfaces with the prototype vehicle. First, we explored the obstacle detection problem. Having a closed property system with ultrasonic sensors, we managed to build an obstacle map with a higher precision level than the build-in product. An increasing limit detection of the ultrasonic sensors was developed using the Structure from Motion technique. This obstacle occupancy information was exploited afterwards in order to solve the detection problem of the navigation corridors. Second, the vehicle localization is addressed. Three localization techniques work for a continuous functioning robustness: the localization by odometry, the localization by occupancy grid map matching and the localization by comparing the current image with the images stored in a database adapted to our needs and improved by computing means. Last, we interested in the vehicle navigation problem. We considered solved the actuator control problem for the tracking of a given trajectory and we concentrated on an admissible trajectory planning. We developed a local path planning technique for avoiding the unmapped obstacles. In order to build the trajectory we used curves of known geometry and we proved that by using clothoides and eventually two circle arches (if maximum steering angle achieved) it is therefore be possible to create trajectories with continuous curves adapted to our situation. We confirmed that using an obstacle map will allow us to predict forehead the possibility to take a specific navigation corridor. Each part of this work was validated in simulation as well as on real data, proving the pertinence of the proposed approaches for the intended application.

Several of our technological breakthroughs are influenced by types of behavior and structures developed in the natural world, including the emulation of swarm in- telligence and the engineering of artificial synapses that function like the human mind. Much like these breakthroughs, this report examines emerging behaviors across swarms of non-communicating, adaptive units that evade obstacles while find- ing a path, to present a swarming algorithm premised on a class of local rule sets re- sulting in a Unmanned Aerial Vehicle (UAV) group navigating together as a unified swarm. Primarily, this method’s important quality is that its rules are local in nature. Thus, the exponential calculations which can be supposed with growing number of drones, their states, and potential tasks are remedied. To this extent, the study tests the algorithmic rules in experiments to replicate the desired behavior in a bounded virtual space filled with simulated units. Simultaneously, in the adaptation of natural flocking rules the study also introduces the rule sets for goal seeking and uncertainty evasion. In effect, the study succeeds in reaching and displaying the desired goals even as the units avoid unknown before flight obstacles and inter-unit collisions with- out the need for a global centralized command nor a leader based hierarchical system.

Small UAVs are used for low altitude surveillance flights where unknown obstacles can be encountered. These UAVs can be given the capability to navigate in uncertain environments if obstacles are identified. This research presents an obstacle avoidance system for small UAVs. First, a mission waypoint path is created that avoids all known obstacles using a genetic algorithm. Then, while the UAV is in flight, obstacles are detected using a forward looking, onboard camera. Image features are found using the Harris Corner Detector and tracked through multiple video frames which provides three dimensional localization of the features. A sparse three dimensional map of features provides a rough estimate of obstacle locations. The features are grouped into potentially hazardous areas. The small UAV then employs a sliding mode control law on the autopilot to avoid obstacles. This research compares rapidly-exploring random trees to genetic algorithms for UAV pre-mission path planning. It also presents two methods for using image feature movement and UAV telemetry to calculate depth and detect obstacles. The first method uses pixel ray intersection and the second calculates depth from image feature movement. Obstacles are avoided with a success rate of 96%.

In recent years, Unmanned aerial vehicles (UAVs) have developed rapidly due to advances in aerospace technology, and wireless communication systems. As a result of their versatility, cost-effectiveness, and flexibility of deployment, UAVs have been developed to accomplish a variety of large and complex tasks without terrain restrictions, such as battlefield operations, search and rescue under disaster conditions, monitoring, etc. Data collection and offloading missions in The internet of thingss (IoTs) networks can be accomplished with the use of UAVs as network edge nodes. The fundamental challenge in such scenarios is to develop a UAV movement policy that enhances the quality of mission completion and avoids collisions. Real-time learning based on neural networks has been proven to be an effective method for solving decision-making problems in a dynamic, unknown environment. In this thesis, we assume a real-life scenario in which a UAV collects data from Ground base stations (GBSs) without knowing the information of the environment. A UAV is responsible for the MOO including collecting data, avoiding obstacles, path planning, and conserving energy. Two Deep reinforcement learnings (DRLs) approaches were implemented in this thesis and compared. / Under de senaste åren har UAV utvecklats snabbt på grund av framsteg inom flygteknik och trådlösa kommunikationssystem. Som ett resultat av deras mångsidighet, kostnadseffektivitet och flexibilitet i utbyggnaden har UAV:er utvecklats för att utföra en mängd stora och komplexa uppgifter utan terrängrestriktioner, såsom slagfältsoperationer, sök och räddning under katastrofförhållanden, övervakning, etc. Data insamlings- och avlastningsuppdrag i IoT-nätverk kan utföras med användning av UAV:er som nätverkskantnoder. Den grundläggande utmaningen i sådana scenarier är att utveckla en UAV-rörelsepolicy som förbättrar kvaliteten på uppdragets slutförande och undviker kollisioner. Realtidsinlärning baserad på neurala nätverk har visat sig vara en effektiv metod för att lösa beslutsfattande problem i en dynamisk, okänd miljö. I den här avhandlingen utgår vi från ett verkligt scenario där en UAV samlar in data från GBS utan att känna till informationen om miljön. En UAV är ansvarig för MOO inklusive insamling av data, undvikande av hinder, vägplanering och energibesparing. Två DRL-metoder implementerades i denna avhandling och jämfördes.

The adoption of small-size Unmanned Aerial Vehicles (UAVs) in the commercial and professional sectors is rapidly growing. The miniaturisation of sensors and processors, the advancements in connected edge intelligence and the exponential interest in Artificial Intelligence (AI) are boosting the affirmation of autonomous nano-size drones in the Internet of Things (IoT) ecosystem. However, achieving safe autonomous navigation and high-level tasks like exploration and surveillance with these tiny platforms is extremely challenging due to their limited resources. Lightweight and reliable solutions to this challenge are subject to ongoing research. This work focuses on enabling the autonomous flight of a pocket-size, 30-gram platform called Crazyflie in a partially known environment. We implement a modular pipeline for the safe navigation of the nano drone between waypoints. In particular, we propose an AI-aided, vision-based reactive planning method for obstacle avoidance. We deal with the constraints of the nano drone by splitting the navigation task into two parts: a deep learning-based object detector runs on external hardware while the planning algorithm is executed onboard. For designing the reactive approach, we take inspiration from existing sensorbased navigation solutions and obtain a novel method for obstacle avoidance that does not rely on distance information. In the study, we also analyse the communication aspect and the latencies involved in edge offloading. Moreover, we share insights into the finetuning of an SSD MobileNet V2 object detector on a custom dataset of low-resolution, grayscale images acquired with the drone. The results show the ability to command the drone at ∼ 8 FPS and a model performance reaching a COCO mAP of 60.8. Field experiments demonstrate the feasibility of the solution with the drone flying at a top speed of 1 m/s while steering away from an obstacle placed in an unknown position and reaching the target destination. Additionally, we study the impact of a parameter determining the strength of the avoidance action and its influence on total path length, traversal time and task completion. The outcome demonstrates the compatibility of the communication delay and the model performance with the requirements of the real-time navigation task and a successful obstacle avoidance rate reaching 100% in the best-case scenario. By exploiting the modularity of the proposed working pipeline, future work could target the improvement of the single parts and aim at a fully onboard implementation of the navigation task, pushing the boundaries of autonomous exploration with nano drones. / Användningen av små obemannade flygfarkoster (UAV) inom den kommersiella och professionella sektorn ökar snabbt. Miniatyriseringen av sensorer och processorer, framstegen inom connected edge intelligence och det exponentiella intresset för artificiell intelligens (AI) ökar användningen av autonoma drönare i nanostorlek i ekosystemet för sakernas internet (IoT). Att uppnå säker autonom navigering och uppgifter på hög nivå, som utforskning och övervakning, med dessa små plattformar är dock extremt utmanande på grund av deras begränsade resurser. Lättviktiga och tillförlitliga lösningar på denna utmaning är föremål för pågående forskning. Detta arbete fokuserar på att möjliggöra autonom flygning av en 30-grams plattform i fickformat som kallas Crazyflie i en delvis känd miljö. Vi implementerar en modulär pipeline för säker navigering av nanodrönaren mellan riktpunkter. I synnerhet föreslår vi en AI-assisterad, visionsbaserad reaktiv planeringsmetod för att undvika hinder. Vi hanterar nanodrönarens begränsningar genom att dela upp navigeringsuppgiften i två delar: en djupinlärningsbaserad objektdetektor körs på extern hårdvara medan planeringsalgoritmen exekveras ombord. För att utforma den reaktiva metoden hämtar vi inspiration från befintliga sensorbaserade navigeringslösningar och tar fram en ny metod för hinderundvikande som inte är beroende av avståndsinformation. I studien analyserar vi även kommunikationsaspekten och de svarstider som är involverade i edge offloading. Dessutom delar vi med oss av insikter om finjusteringen av en SSD MobileNet V2-objektdetektor på en skräddarsydd dataset av lågupplösta gråskalebilder som tagits med drönaren. Resultaten visar förmågan att styra drönaren med ∼ 8 FPS och en modellprestanda som når en COCO mAP på 60.8. Fältexperiment visar att lösningen är genomförbar med drönaren som flyger med en topphastighet på 1 m/s samtidigt som den styr bort från ett hinder som placerats i en okänd position och når måldestinationen. Vi studerar även effekten av en parameter som bestämmer styrkan i undvikandeåtgärden och dess påverkan på den totala väglängden, tidsåtgången och slutförandet av uppgiften. Resultatet visar att kommunikationsfördröjningen och modellens prestanda är kompatibla med kraven för realtidsnavigering och ett lyckat undvikande av hinder som i bästa fall uppgår till 100%. Genom att utnyttja modulariteten i den föreslagna arbetspipelinen kan framtida arbete inriktas på förbättring av de enskilda delarna och syfta till en helt inbyggd implementering av navigeringsuppgiften, vilket flyttar gränserna för autonom utforskning med nano-drönare.

Most micro air vehicles rely heavily on reliable GPS measurements for proper estimation and control, and therefore struggle in GPS-degraded environments. When GPS is not available, the global position and heading of the vehicle is unobservable. This dissertation establishes the theoretical and practical advantages of a relative navigation framework for MAV navigation in GPS-degraded environments. This dissertation explores how the consistency, accuracy, and stability of current navigation approaches degrade during prolonged GPS dropout and in the presence of heading uncertainty. Relative navigation (RN) is presented as an alternative approach that maintains observability by working with respect to a local coordinate frame. RN is compared with several current estimation approaches in a simulation environment and in hardware experiments. While still subject to global drift, RN is shown to produce consistent state estimates and stable control. Estimating relative states requires unique modifications to current estimation approaches. This dissertation further provides a tutorial exposition of the relative multiplicative extended Kalman filter, presenting how to properly ensure observable state estimation while maintaining consistency. The filter is derived using both inertial and body-fixed state definitions and dynamics. Finally, this dissertation presents a series of prolonged flight tests, demonstrating the effectiveness of the relative navigation approach for autonomous GPS-degraded MAV navigation in varied, unknown environments. The system is shown to utilize a variety of vision sensors, work indoors and outdoors, run in real-time with onboard processing, and not require special tuning for particular sensors or environments. Despite leveraging off-the-shelf sensors and algorithms, the flight tests demonstrate stable front-end performance with low drift. The flight tests also demonstrate the onboard generation of a globally consistent, metric, and localized map by identifying and incorporating loop-closure constraints and intermittent GPS measurements. With this map, mission objectives are shown to be autonomously completed.

L'objectif du travail proposé est de concevoir et commander un groupe des robots mobiles similaires et d'architecture simple appelés m-bots (mono-robots). Plusieurs m-bots ont la capacité de saisir ensemble un objet afin d'assurer sa co-manipulation et son transport quelle que soit sa forme et sa masse. Le robot résultant est appelé p-bot (poly-robot) et est capable d'effectuer des tâches de déménageur pour le transport d'objets génériques. La reconfigurabilité du p-bot par l'ajustement du nombre des m-bots utilisés permet de manipuler des objets lourds et des objets de formes quelconques (particulièrement s'ils sont plus larges qu'un seul m-bot). Sont considérés dans ce travail l'évitement d'obstacle ainsi que la stabilité du p-bot incluant la charge à transporter. Une cinématique pour un mécanisme de manipulation a été proposée et étudiée. Ce dernier assure le levage de la charge et son dépôt sur le corps des robots pour la transporter. Plusieurs variantes d'actionnement ont été étudiées : passif, avec compliance et actionné. Un algorithme de positionnement optimal des m-bots autour de l'objet à manipuler a été proposé afin d'assurer la réussite de la tâche à effectuer par les robots. Cet algorithme respecte le critère de "Force Closure Grasping" qui assure la stabilité de la charge durant la phase de manipulation. Il maintient aussi une marge de stabilité statique qui assure la stabilité de l'objet durant la phase de transport. Enfin, l'algorithme respecte le critère des zones inaccessibles qui ne peuvent pas être atteintes par les m-bots. Une loi de commande a été utilisée afin d'atteindre les positions désirées pour les m-bots et d'assurer la navigation en formation, durant la phase du transport, durant laquelle chaque robot élémentaire doit maintenir une position désirée par rapport à l'objet transporté. Des résultats de simulation pour un objet de forme quelconque, décrite par une courbe paramétrique, sont présentés. Des simulations 3D en dynamique multi-corps ainsi que des expériences menées sur les prototypes réalisés ont permis de valider nos propositions. / Our goal in the proposed work is to design and control a group of similar mobile robots with a simple architecture, called m-bot. Several m-bots can grip a payload, in order to co-manipulate and transport it, whatever its shape and mass. The resulting robot is called a p-bot andis capable to solve the so-called "removal-man task" to transport a payload. Reconfiguring the p-bot by adjusting the number of m-bots allows to manipulate heavy objects and to manage objects with anyshape, particularly if they are larger than a single m-bot. Obstacle avoidance is addressed and mechanical stability of the p-bot and its payload is permanently guaranteed. A proposed kinematic architecture for a manipulation mechanism is studied. This mechanism allows to lift a payload and put it on them-bot body in order to be transported. The mobile platform has a free steering motion allowing the system maneuver in any direction. An optimal positioning of the m-bots around the payload ensures a successful task achievement without loss of stability for the overall system. The positioning algorithm respects the Force Closure Grasping (FCG) criterion which ensures the payload stability during the manipulation phase. It respects also the Static Stability Margin (SSM) criterion which guarantees the payload stability during the transport. Finally, it considers also the Restricted Areas (RA) that could not be reached by the robots to grab the payload. A predefined control law is then used to ensure the Target Reaching (TR) phase of each m-bot to its desired position around the payload and to track a Virtual Structure (VS), during the transportation phase, in which each elementary robot has to keep the desired position relative to the payload. Simulation results for an object of any shape, described by aparametric curve, are presented. Additional 3D simulation results with a multi-body dynamic software and experiments by manufactured prototypes validate our proposal.

Micro aerial vehicles and other autonomous systems have the potential to truly transform life as we know it, however much of the potential of autonomous systems remains unrealized because reliable navigation is still an unsolved problem with significant challenges. This dissertation presents solutions to many aspects of autonomous navigation. First, it presents ROSflight, a software and hardware architure that allows for rapid prototyping and experimentation of autonomy algorithms on MAVs with lightweight, efficient flight control. Next, this dissertation presents improvments to the state-of-the-art in optimal control of quadrotors by utilizing the error-state formulation frequently utilized in state estimation. It is shown that performing optimal control directly over the error-state results in a vastly more computationally efficient system than competing methods while also dealing with the non-vector rotation components of the state in a principled way. In addition, real-time robust flight planning is considered with a method to navigate cluttered, potentially unknown scenarios with real-time obstacle avoidance. Robust state estimation is a critical component to reliable operation, and this dissertation focuses on improving the robustness of visual-inertial state estimation in a filtering framework by extending the state-of-the-art to include better modeling and sensor fusion. Further, this dissertation takes concepts from the visual-inertial estimation community and applies it to tightly-coupled GNSS, visual-inertial state estimation. This method is shown to demonstrate significantly more reliable state estimation than visual-inertial or GNSS-inertial state estimation alone in a hardware experiment through a GNSS-GNSS denied transition flying under a building and back out into open sky. Finally, this dissertation explores a novel method to combine measurements from multiple agents into a coherent map. Traditional approaches to this problem attempt to solve for the position of multiple agents at specific times in their trajectories. This dissertation instead attempts to solve this problem in a relative context, resulting in a much more robust approach that is able to handle much greater intial error than traditional approaches.

You do not want your drone to crash. Therefore, safety systems should be put in place to prevent such an event, and obstacle avoidance is a major part of this. Today, the most successful techniques use cameras or light detection and ranging (LIDAR) to find and avoid obstacles; but to improve resiliency, multiple systems should be used. This thesis proposes to use microphones, listening to the drone’s own noise, to estimate the distance to surrounding obstacles. An obstacle echolocation solution for multi-rotor aerial vehicles (MAVs) using ego-noise is developed. The MAV’s noise is captured and auto-correlated to detect echoes at different time delays. This signal is whitened to remove structured measurement noise resulting from the narrow-band components of the MAV’s noise. By recording the MAV’s noise using multiple microphones, a time of arrival (TOA) estimate of the obstacle position is achieved. A beamforming-based solution is used to calculate this estimate. A series of simplified proof-of-concept experiments show that ego-noise echolocation is possible and that the developed solution works in a controlled environment. A prototype implementation of a realistic system is also created. Four signal fusion alternatives are compared, though no best alternative is found for all situations. More work is needed to apply the findings of this work in a robust way, but the principle is shown to work.

<p>This thesis presents solutions to some of the problems facing Autonomous Surface Vehicle (ASV) deployments in inland waterways through the development of navigational and control systems. Fluvial systems are one of the hardest inland waterways to navigate and are thus used as a use-case for system development. The systems are built to reduce the reliance on a-prioris during ASV operation. This is crucial for exceptionally dynamic environments such as fluvial bodies of water that have poorly defined routes and edges, can change course in short time spans, carry away and deposit obstacles, and expose or cover shoals and man-made structures as their water level changes. While navigation of fluvial systems is exceptionally difficult potential autonomous data collection can aid in important scientific missions in under studied environments.</p> <p><br></p> <p>The work has four contributions targeting solutions to four fundamental problems present in fluvial system navigation and control. To sense the course of fluvial systems for navigable path determination a fluvial segmentation study is done and a novel dataset detailed. To enable rapid path computations and augmentations in a fast moving environment a Dubins path generator and augmentation algorithm is presented ans is used in conjunction with an Integral Line-Of-Sight (ILOS) path following method. To rapidly avoid unseen/undetected obstacles present in fluvial environments a Deep Reinforcement Learning (DRL) agent is built and tested across domains to create dynamic local paths that can be rapidly affixed to for collision avoidance. Finally, a custom low-cost and deployable ASV, BREAM (Boat for Robotic Engineering and Applied Machine-Learning), capable of operating in fluvial environments is presented along with an autonomy package used in providing base level sensing and autonomy processing capability to varying platforms.</p> <p><br></p> <p>Each of these contributions form a part of a larger documented Fluvial Navigation Control Architecture (FNCA) that is proposed as a way to aid in a-priori free navigation of fluvial waterways. The architecture relates the navigational structures into high, mid, and low-level controller Guidance and Navigational Control (GNC) layers that are designed to increase cross vehicle and domain deployments. Each component of the architecture is documented, tested, and its application to the control architecture as a whole is reported.</p>