Dissertations / Theses on the topic 'Robot inverse dynamics'

Robotic manipulators have been extensively used in industrial automation, hazardous environments and outer space. The requirement for increased speeds of operation and lightweight design has made structural flexibility the constraining factor in robot design. For current manipulators that exhibit significant link flexibility, the control methodology seems to be to drive them so slowly that link dynamics are not excited. To make significant gains in speed and efficiency, the deflections due to link deformation will have to be compensated. With the ready availability of powerful processors, the cost of implementing complex control methodologies is not excessive. Flexible-link robots provide significant challenges for the control engineer. With nonlinear configuration-dependent and nonminimum-phase dynamics, the control of the end-effector in task-space is one of the most difficult problems encountered. Fortunately, flexible-link robots exhibit one beneficial characteristic. The map from joint torques to joint rate is passive, allowing a strictly passive feedback controller to produce a stable system. The requirement then is to produce a reference joint trajectory that will result in the end-effector following its desired track. This thesis deals with the problem of inverting the nonlinear nonrninimum-phase dynamics to produce a feedforward torque and joint trajectory from a given end-effector trajectory. Generally the inverse dynamics will be noncasual, that is the output of the inverse system will depend on future inputs. In the linear case this noncasual inverse can be solved using Fourier transform methods on the complete trajectory. In this work we have assumed that the complete end-effector trajectory is not available. The input may come from an operator controlling the movement by sight or by a system that is updating the trajectory as it analyses its own sensors. Because of this restriction, the inverse dynamics are approximated by a casual system which only uses past inputs. Using a nonlinear inner-outer factorisation of the dynamics, inverting the outer factor and approximating the inverse of the inner factor with its static inverse, an approximate inverse-dynamics system was generated. Alternatively, by modifying the input of the dynamics to be all of the rigid contribution plus a fraction of the elastic contribution, a stable inverse was generated. Both of these approximate inverses have been implemented on a planar three-DOF system with the first two links flexible. Simulation and implementation on an experimental facility has shown that approximate end-effector tracking can be obtained. While an approximate inverse based feedforward cannot produce perfect tracking, it is a significant improvement over the current standard of generating a joint trajectory based on the inverse of the equivalent rigid robot.

Multiple robot manipulators cooperating in a common manipulation task can accomplish complex tasks that a single manipulator would be unable to complete. To achieve physical cooperation with multiple manipulators working on a common object, interaction forces need to be controlled throughout the motion. The aim of this research is to develop an inverse dynamics model-based cooperative force and position control scheme for multiple robot manipulators. An extended definition of motion is proposed to include force demands based on a constrained Lagrangian dynamics and Lagrangian multipliers formulation. This allows the direct calculation of the inverse dynamics with both motion and force demands. A feedforward controller based on the proposed method is built to realise the cooperative control of two robots sharing a common load, with both motion and force demands. Furthermore, this thesis develops a method to design an optimal excitation trajectory for robot dynamic parameter estimation utilising the Schroeder Phased Harmonic Sequence. This method yields more precise and accurate inverse dynamics models, which result in better control. The proposed controller is then tested in an experimental set-up consisting of two robot manipulators and a common load. Results show that in general the proposed controller performs noticeably better position and force tracking, especially for higher speed motions, when compared to traditional hybrid position/force controllers.

Este trabalho trata do estudo de robôs de arquitetura paralela, focando na modelagem e otimização dos mesmos. Não foi construído nenhum tipo de protótipo físico, contudo os modelos virtuais poderão, no futuro, habilitar tal façanha. Após uma busca por uma aplicação que se beneficie do uso de um robô de arquitetura paralela, fez-se uma pesquisa por arquiteturas viáveis já existentes ou relatadas na literatura. Escolheu-se a mais apta e prosseguiu-se com os estudos e modelagem cinemática e dinâmica, dando uma maior ênfase na cinemática e dinâmica inversa, esta última utilizando a formulação de Newton - Euler. Foi construído um simulador virtual em ambiente MATLAB 6.5, dotado de várias capacidades como interpolação linear e circular, avanço e uso de múltiplos eixos coordenados. Seu propósito principal é o de demonstrar a funcionalidade e eficácia dos métodos utilizados. Depois foi incorporado ao simulador um algoritmo de cálculo do volume de trabalho da máquina que utiliza alguns dados do usuário para calcular o volume, que pode ser aquele atrelado a uma postura em particular ou o volume de trabalho de orientação total. Algoritmos para medir o desempenho da máquina quanto à uniformidade e utilização da força dos atuadores foram construídos e também incorporados ao simulador, que consegue mostrar o elipsóide de forças ao longo de quaisquer movimentos executados pela plataforma móvel. Quanto à otimização, parte do ferramental previamente construído foi utilizado para que se pudesse chegar a um modelo de uma máquina que respeitasse restrições mínimas quanto ao tamanho e forma de seu volume de trabalho, mas ainda mantendo o melhor desempenho possível dentro deste volume.<br>This work is about the study of parallel architecture robots, focusing in modeling and optimization. No physical prototypes were built, although the virtual models can help those willing to do so. After searching for an application that could benefit from the use of a parallel robot, another search was made, this time for the right architecture type. After selecting the architecture, the next step was the kinematics and dynamics analysis. The dynamics model is developed using the Newton ? Euler method. A virtual simulator was also developed in MATLAB 6.5 environment. The simulator?s main purpose was to demonstrate that the methods applied were correct and efficient, so it has several features such as linear and circular interpolations, capacity to use multiple coordinate systems and others. After finishing the simulator, an algorithm to calculate the machine workspace was added. The algorithm receives as input some desired requirements regarding the manipulator pose and then calculates the workspace, taking into consideration imposed constraints. Lastly, algorithms capable to measure the manipulator?s performance regarding to its actuator and end-effector force relationship were also incorporated into the simulator that calculates the machine?s force ellipsoid during any movement, for each desired workspace point. For the optimization procedures, some previously developed tools were used, so that the resulting model was capable to respect some workspace constraints regarding size and shape, but also maintaining the best performance possible inside this volume.

Les grands robots marcheurs sont des systèmes mécatroniques poly-articulés complexes qui cristallisent la volonté des humains de conférer leurs capacités à des artefacts, l’une d’entre elle étant la locomotion bipède, et plus particulièrement la conservation de l’équilibre face à des perturbations extérieures. Cette thèse propose un stabilisateur postural ainsi que sa mise en œuvre sur le système locomoteur BIP 2000.Ce robot anthropomorphique possède quinze degrés de libertés actionnés par moteurs électriques et a reçu un nouvel automate ainsi que des variateurs industriels lors de la mise à jour réalisée dans le cadre de ces travaux. Un contrôleur a été conçu et implémenté en suivant les principes de la programmation orientée objet afin de fournir une modularité qui s’inspire de la symétrie naturelle des humanoïdes. Cet aspect a conduit à l’élaboration d’un ensemble d’outils mathématiques permettant de calculer l’ensemble des modèles d’un robot composé de sous-robots dont on connaîtrait déjà les modèles. Le contrôleur permet notamment à la machine de suivre des trajectoires calculées hors ligne par des algorithmes de génération de marches dynamiques ainsi que de tester le stabilisateur postural.Ce dernier consiste en un contrôle en position du robot physique par la consigne d’un robot virtuel de modèle dégradé, commandé en effort, soumis à des champs électrostatiques contraignant sa configuration articulaire. Les tests effectués ont permis de montrer la faisabilité de la méthode<br>Big walking robots are complex multi-joints mechanical systems which crystallize the human will to confer their capabilities on artefacts, one of them being the bipedal locomotion and more especially the balance keeping against external disturbances. This thesis proposes a balance stabilizer under operating conditions displayed on the locomotor system BIP 2000.This anthropomorphic robot has got fifteen electrically actuated degree of freedom and an Industrial controller. A new software has been developed with an object-oriented programming approach in order to propose the modularity required by the emulated and natural human symmetry. This consideration leads to the development of a mathematical tool allowing the computation of every modelling of a serial robot which is the sum of multiple sub robots with already known modelling. The implemented software also enables the robot to run offline generated dynamic walking trajectories and to test the balance stabilizer.We explore in this thesis the feasibility of controlling the center of gravity of a multibody robotic system with electrostatic fields acting on its virtual counterpart in order to guarantee its balance. Experimental results confirm the potential of the proposed approach

The purpose of this thesis is to develop a position control method for parallel manipulators so that the end effector can follow a desired trajectory specified in the task space where joint flexibility that occurs at the actuated joints is also taken into consideration. At the beginning of the study, a flexible joint is modeled, and the equations of motion of the parallel manipulators are derived for both actuator variables and joint variables by using the Lagrange formulation under three assumptions regarding dynamic coupling between the links and the actuators. These equations of motion are transformed to an input/output relation between the actuator torques and the actuated joint variables to achieve the trajectory tracking control. Moreover, the singular configurations of the parallel manipulators are explained. As a case study, a three degree of freedom, two legged planar parallel manipulator is simulated considering joint flexibility. The structural damping of the active joints, viscous friction at the passive joints and the rotor damping are also considered throughout the study. Matlab&reg<br>and Simulink&reg<br>softwares are used for the simulations. The results of the simulations reveal that steady state errors are negligibly small and good tracking performances can be achieved.

In this thesis, a technique for the motion of parallel manipulators through drive singularities is investigated. To remedy the problem of unbounded inverse dynamics solution in the neighborhood of drive singularities, an inverse dynamics controller which uses a conventional inverse dynamics control law outside the neighborhood of singularities and switches to the mode based on the formerly derived modified equations inside the neighborhood of singularities is proposed. As a result, good tracking performance is obtained while the actuator forces remain within the saturation limits of the actuators around singular configurations.

In the context of control, the motion of a legged robot is very challenging compared with traditional fixed manipulator. Recently, many researches have been conducted to control the motion of legged robot with different techniques. On the other hand, manipulation tasks have been addressed in many applications. These researches solved either the mobility or the manipulation problems, but integrating both properties in one system is still not available. In this thesis, a control algorithm is presented to control both locomotion and manipulation in a six legged robot. Landmines detection process is considered as a case study of this project to accelerate the mine detection operation by performing both walking and scanning simultaneously. In order to qualify the robot to perform more tasks in addition to the walking task, the joint redundancy of the robot is exploited optimally. The tasks are arranged according to their importance to high level of priority and low level of priority. A new task priority redundancy resolution technique is developed to overcome the effect of the algorithmic singularities and the kinematic singularity. The computational aspects of the solution are also considered in view of a real-time implementation. Due to the dynamic changes in the size of the robot motion space, the algorithm has the ability to make a trade-off between the number of achieved tasks and the imposed constraints. Furthermore, an appropriate hierarchy is imposed in order to ensure an accurate decoupling between the executed tasks. The dynamic effect of the arm on the overall performance of the robot is attenuated by reducing the optimisation variables. The effectiveness of the method is evaluated on a Computer Aided Design (CAD) model and the simulations of the whole operation are conducted using MATLAB and SimMechanics.

The focus of this document is on the design, modeling, and control of a self-balancing two wheel robot, hereafter referred to as the balance bot, driven by independent brushless DC (BLDC) motors. The balance bot frame is composed of stacked layers allowing a lightweight, modular, and rigid mechanical design. The robot is actuated by a pair of brushless DC motors equipped with Hall effect sensors and encoders allowing determination of the angle and angular velocity of each wheel. Absolute orientation measurement is accomplished using a full 9-axis IMU consisting of a 3-axis gyroscope, a 3-axis accelerometer, and a 3-axis magnetometer. The control algorithm is designed to minimize deviations from a set point specified by an external radio remote control, which allows the remote operator to steer and drive the bot wirelessly while it remains balanced. Multiple dynamic models are proposed in this analysis, and the selected model is used to develop a linear-quadratic regulator based state-feedback controller to perform reference tracking. Controller tracking performance is improved by incorporating a prefilter stage between the setpoint command from the remote control and the state-feedback controller. Modeling of the actuator dynamics is considered brie y and is discussed in relation to the control algorithm used to balance the robot. Electrical and software design implementations are also presented with a focus on effective implementation of the proposed control algorithms. Simulated and physical testing results show that the proposed balance bot and controller design are not only feasible but effective as a means of achieving robust performance under dynamic tracking profiles provided by the remote control.

Parallel robots with configurable platform are a class of parallel robots in which the end-effector is a closed-loop flexible chain of rigid links. We have developed a 5-RRR planar mechanism that features a flexible 5-bar chain as end-effector. The angles between adjacent sides of this chain can be controlled through the actuated revolute joints attached to the base of the mechanism. This thesis consists in the geometrical design of n-RRR planar parallel robots and in the study of the Direct Kinematics for 4-, 5- and 6-RRR mechanisms using Bilateration, a method that greatly reduces the computational time for the kinematic analysis. The next step is the singularity analysis for the n-RRR robot architectures; finally, in the last part of this thesis we present the results from experimental tests that have been performed on a 5-RRR robot prototype.

In this thesis, dynamic modeling and nonlinear control of autonomous underwater vehicle manipulator systems are presented. Mainly, two types of systems consisting of a 6-DOF AUV equipped with a 6-DOF manipulator subsystem (UVMS) and with an 8-DOF redundant manipulator subsystem (UVRMS) are modeled considering hydrostatic forces and hydrodynamic effects such as added mass, lift, drag and side forces. The shadowing effects of the bodies on each other are introduced when computing the hydrodynamic forces. The system equations of motion are derived recursively using Newton&ndash<br>Euler formulation. The inverse dynamics control algorithms are formulated and trajectory tracking control of the systems is achieved by assigning separate tasks for the end effector of the manipulator and for the underwater vehicle. The proposed inverse dynamics controller utilizes the full nonlinear model of the system and consists of a linearizing control law that uses the feedback of positions and velocities of the joints and the underwater vehicle in order to cancel off the nonlinearities of the system. The PD control is applied after this complicated feedback linearization process yielding second order error dynamics. The thruster dynamics is also incorporated into the control system design. The stability analysis is performed in the presence of parametric uncertainty and disturbing ocean current. The effectiveness of the control methods are demonstrated by simulations for typical underwater missions.

Model-based control strategies for robot manipulators can present numerous performance advantages when an accurate model of the system dynamics is available. In practice, obtaining such a model is a challenging task which involves modeling such physical processes as friction, which may not be well understood and difficult to model. Furthermore, uncertainties in the physical parameters of a system may be introduced from significant discrepancies between the manufacturer data and the actual system. Traditionally, adaptive and robust control strategies have been developed to deal with parametric uncertainty in the dynamic model, but often require knowledge of the structure of the dynamics. Recent approaches to model-based manipulator control involve data-driven learning of the inverse dynamics relationship, eliminating the need for any a-priori knowledge of the system model. Locally Weighted Projection Regression (LWPR) has been proposed for learning the inverse dynamics function of a manipulator. Due to its use of simple local, linear models, LWPR is suitable for online and incremental learning. Although global regression techniques such as Gaussian Process Regression (GPR) have been shown to outperform LWPR in terms of accuracy, due to its heavy computational requirements, GPR has been applied mainly to offline learning of inverse dynamics. More recent efforts in making GPR computationally tractable for real-time control have resulted in several approximations which operate on a select subset, or sparse representation of the entire training data set. Despite the significant advancements that have been made in the area of learning control, there has not been much work in recent years to evaluate these newer regression techniques against traditional model-based control strategies such as adaptive control. Hence, the first portion of this thesis provides a comparison between a fixed model-based control strategy, an adaptive controller and the LWPR-based learning controller. Simulations are carried out in order to evaluate the position and orientation tracking performance of each controller under varied end effector loading, velocities and inaccuracies in the known dynamic parameters. Both the adaptive controller and LWPR controller are shown to have comparable performance in the presence of parametric uncertainty. However, it is shown that the learning controller is unable to generalize well outside of the regions in which it has been trained. Hence, achieving good performance requires significant amounts of training in the anticipated region of operation. In addition to poor generalization performance, most learning controllers commence learning entirely from `scratch,' making no use of any a-priori knowledge which may be available from the well-known rigid body dynamics (RBD) formulation. The second portion of this thesis develops two techniques for online, incremental learning algorithms which incorporate prior knowledge to improve generalization performance. First, prior knowledge is incorporated into the LWPR framework by initializing the local linear models with a first order approximation of the prior information. Second, prior knowledge is incorporated into the mean function of Sparse Online Gaussian Processes (SOGP) and Sparse Pseudo-input Gaussian Processes (SPGP), and a modified version of the algorithm is proposed to allow for online, incremental updates. It is shown that the proposed approaches allow the system to operate well even without any initial training data, and further performance improvement can be achieved with additional online training. Furthermore, it is also shown that even partial knowledge of the system dynamics, for example, only the gravity loading vector, can be used effectively to initialize the learning.

碩士<br>國立虎尾科技大學<br>機械設計工程研究所<br>99<br>Robot has been developed for many years, the application of the Robot is more and more popular. It is gradually somebody in work. Most of the design of Robot for the low degree of freedom, this will limit its use. To improve this drawback, it made anthropomorphic mechanical fingers, but also cause the inverse kinematics is difficult. In this paper, for a humanoid robot finger with nonlinear coupling joints, a approach is proposed to derive the algebraic-elimination solutions of inverse kinematics. First, the given position of fingertip with transform frames and vector-loop obtain two algebraic equation, and using trigonometric terms conversed into an eight-degree polynomial for describe coupling joints. To generate the smooth jerk-limited federate profile and interpolation points, a dynamics-based interpolator with real-time look-ahead (DBLA) algorithm is applied to plane the motion trajectory. Given interpolation points of fingertip, the iterative Newton-Raphson method with a suitable initial value for solving the roots of polynomial. In order to obtain the dynamic equation of robot finger, using Lagrange method and Newton-Euler method derivation and cross match, to understand the dynamic behavior of Robot finger. Finally, using Matlab software simulation and analysis, verify the correctness and feasibility of the theory.

Indiana University-Purdue University Indianapolis (IUPUI)<br>Khakpour, Zahra M.S.M.E., Purdue University, May 2017. Multibody Dynamics Model of A Full Human Body For Simulating Walking, Major Professor: Hazim El-Mounayri. Bipedal robotics is a relatively new research area which is concerned with creating walking robots which have mobility and agility characteristics approaching those of humans. Also, in general, simulation of bipedal walking is important in many other applications such as: design and testing of orthopedic implants; testing human walking rehabilitation strategies and devices; design of equipment and facilities for human/robot use/interaction; design of sports equipment; and improving sports performance & reducing injury. One of the main technical challenges in that bipedal robotics area is developing a walking control strategy which results in a stable and balanced upright walking gait of the robot on level as well as non-level (sloped/rough) terrains. In this thesis the following aspects of the walking control strategy are developed and tested in a high-fidelity multibody dynamics model of a humanoid body model: 1. Kinematic design of a walking gait using cubic Hermite splines to specify the motion of the center of the foot. 2. Inverse kinematics to compute the legs joint angles necessary to generate the walking gait. 3. Inverse dynamics using rotary actuators at the joints with PD (Proportional-Derivative) controllers to control the motion of the leg links. The thee-dimensional multibody dynamics model is built using the DIS (Dynamic Interactions Simulator) code. It consists of 42 rigid bodies representing the legs, hip, spine, ribs, neck, arms, and head. The bodies are connected using 42 revolute joints with a rotational actuator along with a PD controller at each joint. A penalty normal contact force model along with a polygonal contact surface representing the bottom of each foot is used to model contact between the foot and the terrain. Friction is modeled using an asperity-based friction model which approximates Coulomb friction using a variable anchor-point spring in parallel with a velocity dependent friction law. In this thesis, it is assumed in the model that a balance controller already exists to ensure that the walking motion is balanced (i.e. that the robot does not tip over). A multi-body dynamic model of the full human body is developed and the controllers are designed to simulate the walking motion. This includes the design of the geometric model, development of the control system in kinematics approach, and the simulation setup.

Tang Wai Sum.<br>Thesis (M.Phil.)--Chinese University of Hong Kong, 1999.<br>Includes bibliographical references (leaves 68-70).<br>Chapter 1 --- Introduction --- p.1<br>Chapter 1.1 --- Redundant Manipulators --- p.1<br>Chapter 1.2 --- Inverse Kinematics of Robotic Manipulators --- p.2<br>Chapter 1.3 --- Inverse Dynamics of Robotic Manipulators --- p.4<br>Chapter 1.4 --- Redundancy Resolutions of Manipulators --- p.5<br>Chapter 1.5 --- Motivation of Using Neural Networks for these Applications --- p.9<br>Chapter 1.6 --- Previous Work for Redundant Manipulator Inverse Kinematics and Inverse Dynamics Computation by Neural Networks --- p.9<br>Chapter 1.7 --- Advantages of the Proposed Recurrent Neural Networks --- p.11<br>Chapter 1.8 --- Contribution of this work --- p.11<br>Chapter 1.9 --- Organization of this thesis --- p.12<br>Chapter 2 --- Problem Formulations --- p.14<br>Chapter 2.1 --- Constrained Optimization Problems for Inverse Kinematics Com- putation of Redundant Manipulators --- p.14<br>Chapter 2.1.1 --- Primal and Dual Quadratic Programs for Bounded Joint Velocity Minimization --- p.14<br>Chapter 2.1.2 --- Primal and Dual Linear Programs for Infinity-norm Joint Velocity Minimization --- p.15<br>Chapter 2.2 --- Constrained Optimization Problems for Inverse Dynamics Com- putation of Redundant Manipulators --- p.17<br>Chapter 2.2.1 --- Quadratic Program for Unbounded Joint Torque Mini- mization --- p.17<br>Chapter 2.2.2 --- Primal and Dual Quadratic Programs for Bounded Joint Torque Minimization --- p.18<br>Chapter 2.2.3 --- Primal and Dual Linear Programs for Infinity-norm Joint Torque Minimization --- p.19<br>Chapter 3 --- Proposed Recurrent Neural Networks --- p.20<br>Chapter 3.1 --- The Lagrangian Network --- p.21<br>Chapter 3.1.1 --- Optimality Conditions for Unbounded Joint Torque Min- imization --- p.21<br>Chapter 3.1.2 --- Dynamical Equations and Architecture --- p.22<br>Chapter 3.2 --- The Primal-Dual Network 1 --- p.24<br>Chapter 3.2.1 --- Optimality Conditions for Bounded Joint Velocity Min- imization --- p.24<br>Chapter 3.2.2 --- Dynamical Equations and Architecture for Bounded Joint Velocity Minimization --- p.26<br>Chapter 3.2.3 --- Optimality Conditions for Bounded Joint Torque Mini- mization --- p.27<br>Chapter 3.2.4 --- Dynamical Equations and Architecture for Bounded Joint Torque Minimization --- p.28<br>Chapter 3.3 --- The Primal-Dual Network 2 --- p.29<br>Chapter 3.3.1 --- Energy Function for Infinity-norm Joint Velocity Mini- mization Problem --- p.29<br>Chapter 3.3.2 --- Dynamical Equations for Infinity-norm Joint Velocity Minimization --- p.30<br>Chapter 3.3.3 --- Energy Functions for Infinity-norm Joint Torque Mini- mization Problem --- p.32<br>Chapter 3.3.4 --- Dynamical Equations for Infinity-norm Joint Torque Min- imization --- p.32<br>Chapter 3.4 --- Selection of the Positive Scaling Constant --- p.33<br>Chapter 4 --- Stability Analysis of Neural Networks --- p.36<br>Chapter 4.1 --- The Lagrangian Network --- p.36<br>Chapter 4.2 --- The Primal-Dual Network 1 --- p.38<br>Chapter 4.3 --- The Primal-Dual Network 2 --- p.41<br>Chapter 5 --- Simulation Results and Network Complexity --- p.45<br>Chapter 5.1 --- Simulation Results of Inverse Kinematics Computation in Re- dundant Manipulators --- p.45<br>Chapter 5.1.1 --- Bounded Least Squares Joint Velocities Computation Using the Primal-Dual Network 1 --- p.46<br>Chapter 5.1.2 --- Minimum Infinity-norm Joint Velocities Computation Us- ing the Primal-Dual Network 2 --- p.49<br>Chapter 5.2 --- Simulation Results of Inverse Dynamics Computation in Redun- dant Manipulators --- p.51<br>Chapter 5.2.1 --- Minimum Unbounded Joint Torques Computation Using the Lagrangian Network --- p.54<br>Chapter 5.2.2 --- Minimum Bounded Joint Torques Computation Using the Primal-Dual Network 1 --- p.57<br>Chapter 5.2.3 --- Minimum Infinity-norm Joint Torques Computation Us- ing the Primal-Dual Network 2 --- p.59<br>Chapter 5.3 --- Network Complexity Analysis --- p.60<br>Chapter 6 --- Concluding Remarks and Future Work --- p.64<br>Publications Resulted from the Study --- p.66<br>Bibliography --- p.68

Cette thèse propose une solution au problème de la génération de mouvements pour les robots humanoïdes. Le cadre qui est proposé dans cette thèse génère des mouvements corps-complet en utilisant la dynamique inverse avec l'espace des tâches et en satisfaisant toutes les contraintes de contact. La spécification des mouvements se fait à travers objectifs dans l'espace des tâches et la grande redondance du système est gérée avec une pile de tâches où les tâches moins prioritaires sont atteintes seulement si elles n'interfèrent pas avec celles de plus haute priorité. À cette fin, un QP hiérarchique est utilisé, avec l'avantage d'être en mesure de préciser tâches d'égalité ou d'inégalité à tous les niveaux de la hiérarchie. La capacité de traiter plusieurs contacts non-coplanaires est montrée par des mouvements où le robot s'assoit sur une chaise et monte une échelle. Le cadre générique de génération de mouvements est ensuite appliqué à des études de cas à l'aide de HRP-2 et Romeo. Les mouvements complexes et similaires à l'humain sont obtenus en utilisant l'imitation du mouvement humain où le mouvement acquis passe par un processus cinématique et dynamique. Pour faire face à la nature instantanée de la dynamique inverse, un générateur de cycle de marche est utilisé comme entrée pour la pile de tâches qui effectue une correction locale de la position des pieds sur la base des points de contact permettant de marcher sur un terrain accidenté. La vision stéréo est également introduite pour aider dans le processus de marche. Pour une récupération rapide d'équilibre, le capture point est utilisé comme une tâche contrôlée dans une région désirée de l'espace. En outre, la génération de mouvements est présentée pour CHIMP, qui a besoin d'un traitement particulier<br>This thesis aims at providing a solution to the problem of motion generation for humanoid robots. The proposed framework generates whole-body motion using the complete robot dynamics in the task space satisfying contact constraints. This approach is known as operational-space inverse-dynamics control. The specification of the movements is done through objectives in the task space, and the high redundancy of the system is handled with a prioritized stack of tasks where lower priority tasks are only achieved if they do not interfere with higher priority ones. To this end, a hierarchical quadratic program is used, with the advantage of being able to specify tasks as equalities or inequalities at any level of the hierarchy. Motions where the robot sits down in an armchair and climbs a ladder show the capability to handle multiple non-coplanar contacts. The generic motion generation framework is then applied to some case studies using HRP-2 and Romeo. Complex and human-like movements are achieved using human motion imitation where the acquired motion passes through a kinematic and then dynamic retargeting processes. To deal with the instantaneous nature of inverse dynamics, a walking pattern generator is used as an input for the stack of tasks which makes a local correction of the feet position based on the contact points allowing to walk on non-planar surfaces. Visual feedback is also introduced to aid in the walking process. Alternatively, for a fast balance recovery, the capture point is introduced in the framework as a task and it is controlled within a desired region of space. Also, motion generation is presented for CHIMP which is a robot that needs a particular treatment

Recent work has shown humanoid robots with spinal columns, instead of rigid torsos, benefit from both better balance and an increased ability to absorb external impact. Similarly, snake robots have shown promise as a viable option for exploration in confined spaces with limited human access, such as during power plant maintenance. Both spines and snakes, as well as hyper-redundant manipulators, can simplify to a model of a system with multiple links. The multi-link inverted pendulum is a well known benchmark problem in control systems due to its ability to accommodate varying model complexity. Such a system is useful for testing new learning algorithms or laying the foundation for autonomous control of more complex devices such as robotic spines and multi-segmented arms which currently use traditional control methods or are operated by humans. It is often easy to view these systems as single-agent learners due to the high level of interaction among the segments. However, as the number of links in the system increases, the system becomes harder to control. This work replaces the centralized learner with a team of coevolved agents. The use of a multiagent approach allows for control of larger systems. The addition of transfer learning not only increases the learning rate, but also enables the training of larger teams which were previously infeasible due to extended training times. The results presented support these claims by examining neuro-evolutionary control of 3-, 6-, and 12-link systems with nominal conditions as well as with sensor noise, actuator noise, and the addition of more complex physics.<br>Graduation date: 2012

This thesis introduces a new measure of balance for bipedal robotics called the foot placement estimator (FPE). To develop this measure, stability first is defined for a simple biped. A proof of the stability of a simple biped in a controls sense is shown to exist using classical methods for nonlinear systems. With the addition of a contact model, an analytical solution is provided to define the bounds of the region of stability. This provides the basis for the FPE which estimates where the biped must step in order to be stable. By using the FPE in combination with a state machine, complete gait cycles are created without any precalculated trajectories. This includes gait initiation and termination. The bipedal model is then advanced to include more realistic mechanical and environmental models and the FPE approach is verified in a dynamic simulation. From these results, a 5-link, point-foot robot is designed and constructed to provide the final validation that the FPE can be used to provide closed-loop gait control. In addition, this approach is shown to demonstrate significant robustness to external disturbances. Finally, the FPE is shown in experimental results to be an unprecedented estimate of where humans place their feet for walking and jumping, and for stepping in response to an external disturbance.

Wheeled systems are energy efficient on prepared surfaces like roads and tracks. Legged systems are capable of traversing different terrains but can be lossy. At low speeds and on off-road surfaces, legged systems using dynamic walking can be energy efficient. Towards this objective, the dynamics of the walker needs to be modelled and controlled. In addition, the braking and ground impact losses need to be minimized. This thesis presents analysis and experiments on the dynamics and control of a rimless-spoked-wheel based mobile robot (Chatur ∗) that belongs to a category between wheeled and legged systems. This rolling rimless wheel is effectively a 2D dynamic walker that serves as a platform for investigating the dynamics and energetics of inverted pendulum walking with constant step angle. A pulsed actuation torque is proposed for the system resulting in four torque regimes defined by the ratio of energy losses to available actuator torque. Five physical constraints that impose fundamental limits on the choice of operating points of a generic inverted pendulum walker are expounded and a method for locating optimal operating points is discussed. Chatur’s hardware design is elaborated and a control topology is proposed for pulsed actuation of the dual brushless dc (BLDC) motor driven platform with wheel synchronization. Various actuator torque profiles can be used to achieve dynamic ‘walking’ in a hub-actuated rimless wheel. The proposed pulsed actuation torque gives rise to four torque regimes that achieve sustained walking and a fifth regime where the walker keeps slowing down with each step. The regimes can be identified based on the fraction of stance phase for which the actuator is energized. Theoretical analysis and experimental results are presented. A simple closed-form analytical solution, using hyperbolic functions, is proposed for the stance phase inverted pendulum dynamics considering planar motion. Ground impacts are assumed to cause abrupt drop in velocity. A constant braking torque that lumps together the effect of several loss phenomena is also considered. Based on whether the CoM is rising or falling and whether or not there is an actuating torque, a stance phase can have four types of sub-phases — actuated rise, unactuated rise, actuated fall, unactuated fall. These are concatenated in four different ways to form repeating cycles yielding the four regimes. The experimental set-up is a fixed step-angle walker constructed using two synchronized adjacent rimless wheels independently actuated at the hub. Varying the magnitude and duty ratio of the torque pulse, the four proposed regimes are experimentally shown. The mechanical power consumption and cost of transport are computed from measured motor currents for different average forward speeds. Videos of the walks are also taken. The space of operating points for an inverted pendulum based bipedal dynamic walker in terms of constraints and optimality is investigated. The operating point of the walker can be specified by the combination of initial mid-stance velocity (v0) and step angle (φm) chosen for a given walk. Not all operating points lead to a realizable steady-state gait. Using basic mechanics, a framework of physical constraints that limit the choice of operating points is proposed. The constraint lines thus obtained delimit the valid region of operation of the walker in the v0–φm plane. Within this allowable region, sub-regions that result in various regimes of walking are identified. A given average forward velocity vx,avg can be achieved by several combinations of v0 and φm. Only one of these combinations results in the minimum mechanical power consumption and can be considered the opti-mum operating point for the given vx,avg. A method is proposed for obtaining this optimal operating point based on tangency of the power and velocity contours. Putting together all such operating points for various vx,avg, a family of optimum operating points, called the optimal locus, is obtained. For the energy loss and internal energy models chosen, the optimal locus obtained has a largely constant step angle with increasing speed but tapers off at non-dimensional speeds close to unity. Thus, choosing the right step angle and keeping it fixed over a broad range of speeds could lead to an inverted pendulum walker that is close to optimal from a mechanical energy perspective. The complete hardware design for Chatur and the caveats associated with reliable performance of the mechanical and electrical subsystems are elaborated. In order to en-sure lateral stability, the system uses two contralateral wheels each driven by a separate BLDC hub motor. From a motor drive perspective, the mechanical load belongs to a unique class of dynamic loads whose reflected torque has a characteristic cyclic varia-tion that repeats several times within a mechanical revolution. The proposed control topology has two hierarchical levels, an inner loop for torque control of BLDC motor implemented using a standard proportional-integral controller, and an outer loop for torque reference generation that uses the information on the ground impact instants and the motor position feedback. Ground impacts of the spokes are detected by an accelerometer to initiate the application of torque. The torque pulse magnitude can be set internally or by a manual operator via radio control. The pulse duration is programmable and enables attainment of various torque regimes at different steady state speeds. The wheels are synchronized so that corresponding spokes on both wheels move in unison. This is achieved by including a wheel synchronization loop that compensates for any lag between the wheels. Lag is detected based on number of sector changes in the hall-effect position sensor data received from both motors. An improved BLDC motor drive is developed wherein non-commutating current feedback is used to reduce current spikes during sector transitions. Experimental waveforms for controller validation are shown.