Tails are frequently utilized by animals to enhance their motion agility, dexterity, and versatility, such as a cheetah using its tail to change its body orientation while its legs are all off the ground and a monkey using its tail to stabilize its locomotion on branches. However, limited by technology and application scenarios, most existing legged robots do not include a robotic tail on board. This research aims to explore the possibilities of adding this missing part on legged robots and investigate the tail's functionalities on enhancing the agility, dexterity, and versatility of legged locomotion. In particular, this research focuses on animal-like serpentine tail structure, due to its larger workspace and higher dexterity. The overall research approach consists of two branches: a theoretical branch that focuses on dynamic modeling, analysis, and control of the legged robots with a serpentine robotic tail; and an empirical branch that focuses on hardware development and experiments of novel serpentine robotic tails and novel legged robots with tail. More specifically, the theoretical work includes modeling and control of a general quadruped platform and a general biped platform, equipped with one of the two general serpentine tail structures: an articulated-structure tail or a continuum-structure tail. Virtual work principle-based formulation was used to formulate the dynamic model. Both classic feedback linearization-based control and optimization-based control were used to coordinate the leg motions and the tail motion. Comparative studies on different tail structures as well as numerical analyses on robotic locomotion were performed to investigate the dynamic effects of serpentine robotic tails. The empirical work includes the developments and experiments of two novel serpentine robotic tail mechanisms and one first-of-its-kind quadruped robot ("VT Lemur") equipped with a serpentine robotic tail. To develop these novel robots, a systematic approach based on dynamic analysis was used. Various experiments were then conducted using the robot hardware. Both the theoretical and empirical results showed that the serpentine robotic tail has significant effects on enhancing the agility, dexterity, and versatility of legged robot motion. / Doctor of Philosophy / Quadruped robots have made impressive progresses over the past decade and now can easily achieve complicated, highly dynamic motions, such as the backflip of the MIT Mini Cheetah robot and the gymnastic parkour motions of the Atlas robot from Boston Dynamics, Inc. However, by looking at nature, many animals use tails to achieve highly agile and dexterous motions. For instance, monkeys are observed to use their tails to grasp branches and to balance their bodies during walking. Kangaroos are found to use their tails as additional limbs to propel and assist their locomotion. Cheetahs and kangaroo rats are thought to use their tails to help maneuvering. Therefore, this research aims to understand the fundamental principles behind these biological observations and develop novel legged robots equipped with a serpentine robotic tail. More specifically, this research aims to answer three key questions: (1) what are the functional benefits of adding a serpentine robotic tail to assist legged locomotion, (2) how do animals control their tail motion, and (3) how could we learn from these findings and enhance the agility, dexterity, and versatility of existing legged robots. To answer these questions, both theoretical investigations and experimental hardware testing were performed. The theoretical work establishes general dynamic models of legged robots with either an articulated tail or a continuum tail. A corresponding motion control framework was also developed to coordinate the leg and tail motions. To verify the proposed theoretical framework, a novel quadruped robot with a serpentine robotic tail was developed and tested.

<p dir="ltr">Motion planning for Autonomous Vehicles (AVs) and Connected Autonomous Vehicles (CAVs) involves the crucial task of translating road environmental data obtained from sensors and connectivity devices into a sequence of executable vehicle actions. This task is critical for AVs and CAVs, because the efficacy of their driving decisions and overall performance depend on the quality of motion planning.</p><p dir="ltr">In the context of motion planning technologies, several fundamental questions and challenges remain despite the widespread adoption of advanced learning-based methods, including deep learning (DL) and deep reinforcement learning (DRL). In this regard, the following critical questions need to be answered: 1) How to design suitable DL architectures to comprehensively understand the driving scenario by integrating data from diverse sources including sensors and connectivity devices? 2) How to effectively use the fused information to make improved driving decisions, accounting for various optimality criteria? 3) How to leverage vehicle connectivity to generate cooperative decisions for multiple CAVs, in a manner that optimizes system-wide utility? 4) How to address the inherent interpretability limitations of DL-based methods to enhance user trust in AVs and CAVs? 5) Is it possible to extend learning-based approaches to operational-level decisions in a way that overcomes the inherent disadvantage of low explainability and lack of safety guarantee?</p><p dir="ltr">In an effort to address these questions and expand the existing knowledge in this domain, this dissertation introduces several learning-based motion planning frameworks tailored towards different driving scenarios of AV and CAV. Technically, these efforts target on developing trustworthy AI systems with a focus on the information fusion, “explainable AI” or XAI and safety critical AI. From a computational perspective, these frameworks introduce new learning-based models with state-of-the-art (SOTA) structures, including Convolutional Neural Network (CNN). Recurrent Neural Networks (RNN), Graph Neural Networks (GNN), Attention networks, and Transformers. They also incorporate reinforcement learning (RL) agents, such as Deep Q Networks (DQN) and Model-based RL. From an application standpoint, these developed frameworks can be deployed directly in AVs and CAVs at Level 3 and above. This can enhance the AV/CAV performance in terms of individual and system performance metrics, including safety, mobility, efficiency, and driving comfort.</p>

Motion planning for robot operations is concerned with path planning and trajectory generation. In multi-robot systems, i.e. with multiple robots operating simultaneously in a shared workspace, the motion planning also needs to coordinate the robots' motions to avoid collisions between them. The multi-robot coordination decides the cycle-time for the planned paths and trajectories since it determines to which extend the operations can take place simultaneously without colliding. To obtain the quickest cycle-time, there needs to bean optimal balance between, on the one hand short paths and fast trajectories, and on the other hand possibly longer paths and slower trajectories to allow that the operations take place simultaneously in the shared workspace. Due to the inter-dependencies, it becomes necessary to consider the path planning, trajectory generation and multi-robot coordination together as one optimisation problem in order to find this optimal balance.This thesis focusses on optimising the motion planning for multi-robot material handling systems of sheet metal parts. A methodology to model the relevant aspects of this motion planning problem together as one multi-disciplinary optimisation problem for Simulation based Optimisation (SBO) is proposed. The identified relevant aspects include path planning,trajectory generation, multi-robot coordination, collision-avoidance, motion smoothness, end-effectors' holding force, cycle-time, robot wear, energy efficiency, part deformations, induced stresses in the part, and end-effectors' design. The cycle-time is not always the (only) objective since it is sometimes equally/more important to minimise robot wear, energy consumption, and/or part deformations. Different scenarios for these other objectives are therefore also investigated. Specialised single- and multi-objective algorithms are proposed for optimising the motion planning of these multi-robot systems. This thesis also investigates how to optimise the velocity and acceleration profiles of the coordinated trajectories for multi-robot material handling of sheet metal parts. Another modelling methodology is proposed that is based on a novel mathematical model that parametrises the velocity and acceleration profiles of the trajectories, while including the relevant aspects of the motion planning problem excluding the path planning since the paths are now predefined.This enables generating optimised trajectories that have tailored velocity and acceleration profiles for the specific material handling operations in order to minimise the cycle-time,energy consumption, or deformations of the handled parts.The proposed methodologies are evaluated in different scenarios. This is done for real world industrial case studies that consider the multi-robot material handling of a multi-stage tandem sheet metal press line, which is used in the automotive industry to produce the cars' body panels. The optimisation results show that significant improvements can be obtained compared to the current industrial practice.

La compréhension et la prédiction des relations structure-fonction de protéines par des approches in sillico représentent aujourd'hui un challenge. Malgré le développement récent de méthodes algorithmiques pour l'étude du mouvement et des interactions moléculaires, la flexibilité de macromolécules reste largement hors de portée des outils actuels de modélisation moléculaire. L'objectif de cette thèse est de développer une nouvelle approche basée sur des algorithmes de planification de mouvement issus de la robotique pour mieux traiter la flexibilité moléculaire dans l'étude des interactions protéiques. Nous avons étendu un algorithme récent d'exploration par échantillonnage aléatoire, ML-RRT pour le désassemblage d'objets articulés complexes. Cet algorithme repose sur la décomposition des paramètres de configuration en deux sous-ensembles actifs et passifs, qui sont traités de manière découplée. Les extensions proposées permettent de considérer plusieurs degrés de mobilité pour la partie passive, qui peut être poussée ou attirée par la partie active. Cet outil algorithmique a été appliqué avec succès pour l'étude des changements conformationnels de protéines induits lors de la diffusion d'un ligand. A partir de cette extension, nous avons développé une nouvelle méthode pour la résolution simultanée du séquençage et des mouvements de désassemblage entre plusieurs objets. La méthode, nommée Iterative-ML-RRT, calcule non seulement les trajectoires permettant d'extraire toutes les pièces d'un objet complexe assemblé, mais également l'ordre permettant le désassemblage. L'approche est générale et a été appliquée pour l'étude du processus de dissociation de complexes macromoléculaires en introduisant une fonction d'évaluation basée sur l'énergie d'interaction. Les résultats présentés dans cette thèse montrent non seulement l'efficacité mais aussi la généralité des algorithmes proposés. / Understanding and predicting structure-function relationships in proteins with fully in silico approaches remain today a great challenge. Despite recent developments of computational methods for studying molecular motions and interactions, dealing with macromolecular flexibility largely remains out of reach of the existing molecular modeling tools. The aim of this thesis is to develop a novel approach based on motion planning algorithms originating from robotics to better deal with macromolecular flexibility in protein interaction studies. We have extended a recent sampling-based algorithm, ML-RRT, for (dis)-assembly path planning of complex articulated objects. This algorithm is based on a partition of the configuration parameters into active and passive subsets, which are then treated in a decoupled manner. The presented extensions permit to consider different levels of mobility for the passive parts that can be pushed or pulled by the motion of active parts. This algorithmic tool is successfully applied to study protein conformational changes induced by the diffusion of a ligand inside it. Building on the extension of ML-RRT, we have developed a novel method for simultaneously (dis)assembly sequencing and path planning. The new method, called Iterative-ML-RRT, computes not only the paths for extracting all the parts from a complex assembled object, but also the preferred order that the disassembly process has to follow. We have applied this general approach for studying disassembly pathways of macromolecular complexes considering a scoring function based on the interaction energy. The results described in this thesis prove not only the efficacy but also the generality of the proposed algorithms

Dans cette thèse nous nous intéressons à deux types d'approches pour la planification de pas pour robots humanoïdes : d'une part les approches discrètes où le robot n'a qu'un nombre fini de pas possibles, et d'autre part les approches où le robot se base sur des zones de faisabilité continues. Nous étudions ces problèmes à la fois du point de vue théorique et pratique. En particulier nous décrivons deux méthodes originales, cohérentes et efficaces pour la planification de pas, l'une dans le cas discret (chapitre 5) et l'autre dans le cas continu (chapitre 6). Nous validons ces méthodes en simulation ainsi qu'avec plusieurs expériences sur le robot HRP-2. / In this thesis we investigate two types of approaches for footstep planning for humanoid robots: on one hand the discrete approaches where the robot has only a finite set of possible steps, and on the other hand the approaches where the robot uses continuous feasibility regions. We study these problems both on a theoretical and practical level. In particular, we describe two original, coherent and efficient methods for footstep planning, one in the discrete case (chapter 5), and one in the continuous case (chapter 6). We validate these methods in simulation and with several experiments on the robot HRP-2.

This thesis work contains a stabilization and a motion planning strategy for a truck and trailer system. A dynamical model for a general 2-trailer with two rigid free joints and a kingpin hitching has been derived based on previous work. The model holds under the assumption of rolling without slipping of the wheels and has been used for control design and as a steering function in a probabilistic motion planning algorithm. A gain scheduled Linear Quadratic (LQ) controller with a Pure pursuit path following algorithm has been designed to stabilize the system around a given reference path. The LQ controller is only used in backward motion and the Pure pursuit controller is split into two parts which are chosen depending on the direction of motion. A motion planning algorithm called Closed-Loop Rapidly-exploring Random Tree (CL-RRT) has then been used to plan suitable reference paths for the system from an initial state configuration to a desired goal configuration with obstacle-imposed constraints. The motion planning algorithm solves a non-convex optimal control problem by randomly exploring the input space to the closed-loop system by performing forward simulations of the closed-loop system. Evaluations of performance is partly done in simulations and partly on a Lego platform consisting of a small-scale system. The controllers have been used on the Lego platform with successful results. When the reference path is chosen as a smooth function the closed-loop system is able to follow the desired path in forward and backward motion with a small control error. In the work, it is shown how the CL-RRT algorithm is able to plan non-trivial maneuvers in simulations by combining forward and backward motion. Beyond simulations, the algorithm has also been used for planning in open-loop for the Lego platform. / <p>Links to movies:</p><p>Reference tracking on Lego platform:</p><p>https://www.dropbox.com/s/ebtfgfo7aq9ij8w/reference_tracking.m4v?dl=0</p><p></p><p>Motion planning simulation with CL-RRT:</p><p>https://www.dropbox.com/s/z9kk27cxdxc1llp/CL_RRT_motion_planning.wmv?dl=0</p>

L’utilisation de robots collaboratifs dans l’industrie de production est en plein essor. Ces robots, dont la puissance est limitée, sont dotés de capteurs leur permettant de détecter la présence d’obstacles, afin de garantir la sécurité des humains se trouvant aux alentours. On s’intéresse dans cette thèse à l’utilisation de systèmes redondants, collaboratifs et mobiles (bras articulés montés sur plateformes mobiles), dans un environnement de production aéronautique peuplé d’humains, pour la réalisation d’opérations d’assemblage. Du point de vue des process, l’utilisation de ces systèmes, souvent beaucoup moins imposants et rigides que leurs homologues non collaboratifs, est jalonnée de défis. La grande souplesse mécanique et les faibles couples qui les caractérisent peuvent induire des imprécisions de positionnement et une incapacité à soutenir l’intensité d’une interaction physique. Ce contexte induit également un besoin d’autonomie de ces systèmes, qui sont amenés à travailler dans des environnements en perpétuelle évolution. Dans cette thèse, une formulation de la redondance cinématique est d’abord présentée. Le formalisme associé permet de simplifier l’exploitation de la liberté que ces systèmes possèdent sur le choix des postures à utiliser pour réaliser des tâches de placement statique de l’effecteur. Ce formalisme est ensuite exploité pour améliorer et caractériser le comportement en déformation et la capacité d’application d’efforts des systèmes redondants sériels. Enfin, le sujet de la planification des mouvements de systèmes robotisés dans un environnement dynamique et encombré est considéré. La solution présentée adapte l’algorithme bien connu des Probabilistic RoadMaps pour y inclure une anticipation des trajectoires des obstacles dynamiques. Cette solution permet de planifier des mouvements sécuritaires, peu intrusifs et efficaces, jusqu’à la destination. / Industrial applications involving collaborative robots are regarded with a growing interest. These power-limited systems are embedded with additional sensing capabilities, which allow them to safely work around humans and conquer new industrial grounds. The subject of managing redundant, collaborative and mobile systems, for assembly operations within a human-populated aircraft production environment, is addressed in this thesis. From a process perspective, the use of these smaller and less stiff counterparts of the non-collaborative robots comes with new challenges. Their high mechanical flexibility and weak actuation can cause shortcomings in positioning accuracy or for interaction force sustainment. The ever-changing nature of human-populated environments also requires highly autonomous solutions. In this thesis, a formulation of positional redundancy is presented. It aims at simplifying the exploitation of the freedom redundant manipulators have on static-task-fulfilling postures. The associated formalism is then exploited to characterise and improve the deformational behaviour and the force capacity of redundant serial systems. Finally, the subject of planning motions within cluttered and dynamic environments is addressed. An adaptation of the well-known Probabilistic RoadMaps method is presented – to which obstacles trajectories anticipation has been included. This solution allows to plan safe, efficient and non-intrusive motions to a given destination.

Cette thèse porte sur le développement de lois de commande pour la marche desrobots bipèdes. Le sous actionnement engendré par le basculement, volontaire ouinvolontaire, du pied en appui sur le sol représente une difficulté majeure. Nousabordons ce problème par l’étude de robots plans avec pieds ponctuels.La première partie de la thèse est une compilation des informations issuesde la littérature que nous avons jugées intéressantes. Nous traitons dans unpremier temps de la modélisation adoptée, puis effectuons une revue des différentesméthodes existantes, et présentons la mise en oeuvre expérimentale de l’une d’entre elle : la méthode HZD.Dans une deuxième partie, nous procédons à une étude de la dissipation relativede l’énergie cinétique du robot lorsque le pied impacte le sol. Nous utilisons les résultats issus de cette étude pour planifier des trajectoires de marche dissipant peu d’énergie. De telles trajectoires ont a priori le mérite de préserver la structure du robot et de générer moins de bruit. A contrario, des trajectoires dissipant la majorité de l’énergie du robot sont utilisées pour un arrêt rapide. Une étude numérique a montré que ces résultats sont robustes à des incertitudes de modèle.Enfin, dans une dernière partie, afin de compenser les difficultés liées au sousactionnement, nous proposons d’utiliser le degré de liberté supplémentaire offert par un changement de l’échelle de temps dans les équations de la dynamique (Time Scaling) pour la classe de robots considérée. En utilisant par ailleurs un changement de coordonnées et de feedback, nous dérivons de nouvelles formes normales exactes et approximatives. / This thesis addresses the general problem of the walking control of biped robots. The foot of the robot in contact with the ground may tip over and cause the robot to be undercatuated. This is a major difficulty in term of control. This problem is addressed by considering planar biped robots with point feet.In a first part, we present a standard way of modeling such systems, a litterature review of the existing methods, and then report experimental results of the walking control of a biped robot using the HZD method.In a second part, we perform an analytic and numeric study of the relativekinetic energy dissipation when the foot of the robot impacts the ground. Usingthis study, we design trajectories with low energy dissipation at impact, which a priori result in gaits preserving the hardware of the robot and causing less noise. On the contrary, trajectories dissipating almost all the kinetic energy are used to quickly stop the robot.Finally, in an attempt to alleviate the burden due to underactuation, we proposeto investigate the additional degree of freedom provided, in the control design, by a change of time scale in the dynamic equations (Time-Scaling) for the considered class of biped robots. Using feedback transformations, we derive new exact and approximative normal forms.

Planejamento de movimento tem o propósito de determinar quais movimentos o robô deve realizar para que alcance posições ou configurações desejadas no ambiente sem que ocorram colisões com obstáculos. É comum na robótica móvel simplificar o planejamento de movimento representando o robô pelas coordenadas do seu centro e desconsiderando qualquer restrição cinemática e dinâmica de movimento. Entretanto, a maioria dos robôs móveis possuem restrições cinemáticas não-holonômicas, e para algumas tarefas e robôs, é importante considerar tais restrições juntamente com o modelo dinâmico do robô na tarefa de planejamento. Assim é possível determinar um caminho que possa ser de fato seguido pelo robô. Nesse trabalho é proposto um método de planejamento cinemático-dinâmico que permite planejar trajetórias para robôs móveis usando malhas de estados. Essa abordagem considera a cinemática e a dinâmica do robô para gerar trajetórias possíveis de serem executadas e livre de colisões com obstáculos. Quando obstáculos não representados no mapa são detectados pelos sensores do robô, uma nova trajetória é gerada para desviar desses obstáculos. O planejamento de movimento utilizando malhas de estados foi associado a um algoritmo de desvio de obstáculos baseado no método da janela dinâmica (DWA). Esse método é responsável pelo controle de seguimento de trajetória, garantindo a segurança na realização da tarefa durante a navegação. As trajetórias planejadas foram executadas em duas plataformas distintas. Essas plataformas foram utilizadas em tarefas de navegação em ambientes simulados interno e externo e em ambientes reais. Para navegação em ambientes internos utilizou-se o robô móvel Pioneer 3AT e para navegação em ambientes externos utilizou-se o veículo autônomo elétrico CaRINA 1 que está sendo desenvolvido no ICMC-USP com apoio do Instituto Nacional de Ciência e Tecnologia em Sistemas Embarcados Críticos (INCT-SEC). / Motion planning aims to determine which movements the robot must accomplish to reach a desired position or configuration in the environment without the occurrence of collisions with obstacles. It is common in mobile robotics to simplify the motion planning representing the robot by the coordinates of its center of gravity and ignoring any kinematic and dynamic constraint motion. However, most mobile robots have non-holonomic kinematic constraints, and for some tasks and robots, it is important to consider these constraints together with the dynamic model of the robot in task planning. Thus it is possible to determine a path that can actually be followed by the robot. Here we propose a method for kinematic-dynamic path planning using lattice states. This approach considers the kinematic and dynamic of the robot to generate generate feasible trajectories free of collisions with obstacles. When obstacles not represented on the map are detected by the sensors of the robot, a new trajectory is generated to avoid these obstacles. The motion planning using lattice state was associated with an obstacle avoidance algorithm based on the dynamic window approach (DWA). This method is responsible for trajectory tracking to ensure safety in navigation tasks. This method was applied in two distinct platforms. These platforms were used for navigation tasks in both indoor and outdoor simulated environments, as well as, in real environments. For navigation in indoor environments we used a Pioneer 3AT robot and for outdoor navigation we used the autonomous electric vehicle CaRINA1 being developed at ICMC-USP with support National Institute of Science and Technology in Critical Embedded Systems (INCT-SEC).

Un robot humanoïde est un système complexe doté de nombreux degrés de liberté, et dont le comportement est sujet aux équations non linéaires du mouvement. Par conséquent, la planification de mouvement pour un tel système est une tâche difficile d'un point de vue calculatoire. Dans ce mémoire, nous avons pour objectif de développer une méthode permettant d'utiliser la puissance de calcul des GPUs dans le contexte de la planification de mouvement corps-complet basée sur de l'optimisation. Nous montrons dans un premier temps les propriétés du problème d'optimisation, et des pistes d'étude pour la parallélisation de ce dernier. Ensuite, nous présentons notre approche du calcul de la dynamique, adaptée aux architectures de calcul parallèle. Cela nous permet de proposer une implémentation de notre problème de planification de mouvement sur GPU: contraintes et gradients sont calculés en parallèle, tandis que la résolution du problème même se déroule sur le CPU. Nous proposons en outre une nouvelle paramétrisation des forces de contact adaptée à notre problème d'optimisation. Enfin, nous étudions l'extension de notre travail au contrôle prédictif. / A humanoid robot is a complex system with numerous degrees of freedom, whose behavior is subject to the nonlinear equations of motion. As a result, planning its motion is a difficult task from a computational perspective.In this thesis, we aim at developing a method that can leverage the computing power of GPUs in the context of optimization-based whole-body motion planning. We first exhibit the properties of the optimization problem, and show that several avenues can be exploited in the context of parallel computing. Then, we present our approach of the dynamics computation, suitable for highly-parallel processing architectures. Next, we propose a many-core GPU implementation of the motion planning problem. Our approach computes the constraints and their gradients in parallel, and feeds the result to a nonlinear optimization solver running on the CPU. Because each constraint and its gradient can be evaluated independently for each time interval, we end up with a highly parallelizable problem that can take advantage of GPUs. We also propose a new parametrization of contact forces adapted to our optimization problem. Finally, we investigate the extension of our work to model predictive control.