Bipedal locomotion is a challenging task in the sense that it requires to maintain dynamic balance while steering the gait in potentially complex environments. In this thesis, we take inspiration from the impressive human walking capabilities to design neuromuscular controllers for humanoid robots. More precisely, we control the robot motors to reproduce the action of virtual muscles commanded by neural signals, similarly to what is done during human locomotion. Gait richness and robustness are two key aspects of this work. In other words, the gaits developed in this work can be steered by an external operator, while being resistant to external perturbations. In the beginning of this thesis, we adapt and port an existing reflex-based neuromuscular model from a 2D simulation environment to a real robotic platform. Starting from this model, we progressively iterate and update the neural commands to add new features. The 2D walker controllers are first incremented to generate walking gaits across a range of forward speeds close to the normal human one. By using a similar control method, we also obtain 2D running gaits whose speed can be controlled by a human operator. The walking controllers are later extended to 3D scenarios (i.e. no motion constraint) with the capability to adapt both the forward speed and the heading direction (including curvature). In parallel, we also develop a method to automatically learn neural networks for a given task and we study how flexible feet affect the gait in terms of robustness and energy efficiency.

Nowadays, very few humanoid robots manage to travel in our daily environments. This is mainly due to their limited locomotion capabilities, far from the human ones. Recently, we developed a bio-inspired torque-based controller recruiting virtual muscles driven by reflexes and a central pattern generator. Straight walking experiments were obtained in a 3D simulation environment, resulting in the emergence of human-like and robust gait patterns, with speed modulation capabilities. In this paper, we extend this model, in order to control the steering direction and curvature. Based on human turning strategies, new control pathways are introduced and optimized to reach the sharpest possible turns. In sum, tele-operated motions can be achieved through the control of two scalar inputs (i.e. forward speed and heading). This is particularly relevant for steering the robot on-line, and navigating in cluttered environments. Finally, the biped demonstrated significant robustness during blind walking experiments.

Despite all the effort devoted to generating locomotion algorithms for bipedal walkers, robots are still far from reaching the impressive human walking capabilities, for instance regarding robustness and energy consumption. In this paper, we have developed a bio-inspired torque-based controller supporting the emergence of a new generation of robust and energy-efficient walkers. It recruits virtual muscles driven by reflexes and a central pattern generator, and thus requires no computationally intensive inverse kinematics or dynamics modeling. This controller is capable of generating energy-efficient and human-like gaits (both regarding kinematics and dynamics) across a large range of forward speeds, in a 3D environment. After a single off-line optimization process, the forward speed can be continuously commanded within this range by changing high-level parameters, as linear or quadratic functions of the target speed. Sharp speed transitions can then be achieved with no additional tuning, resulting in immediate adaptations of the step length and frequency. In this paper, we particularly embodied this controller on a simulated version of COMAN, a 95 cm tall humanoid robot. We reached forward speed modulations between 0.4 and 0.9 m/s. This covers normal human walking speeds once scaled to the robot size. Finally, the walker demonstrated significant robustness against a large spectrum of unpredicted perturbations: facing external pushes or walking on altered environments, such as stairs, slopes, and irregular ground.

We present a multibody simulator being used for compliant humanoid robot modelling and report our reasoning for choosing the settings of the simulator’s key features. First, we provide a study on how the numerical integration speed and accuracy depend on the coordinate representation of the multibody system. This choice is particularly critical for mechanisms with long serial chains (e.g. legs and arms). Our second contribution is a full electromechanical model of the inner dynamics of the compliant actuators embedded in the COMAN robot, since joints’ compliance is needed for the robot safety and energy efficiency. Third, we discuss the different approaches for modelling contacts and selecting an appropriate contact library. The recommended solution is to couple our simulator with an open-source contact library offering both accurate and fast contact modelling. The simulator performances are assessed by two different tasks involving contacts: a bimanual manipulation task and a squatting tasks. The former shows reliability of the simulator. For the latter, we report a comparison between the robot behaviour as predicted by our simulation environment, and the real one.

Navigation of humanoids in cluttered environments is a complex task which requires sensorimotor coordination while maintaining the balance of the walker. Typically, robots rely on greedy computation of slow, inefficient and unnatural gaits. This contrasts with the relative ease and efficiency characterizing motion planning and execution of humans. In previous contributions, we developed a bio-inspired torque-based controller recruiting virtual muscles driven by reflexes and a central pattern generator. Speed control and steering could be achieved by the modulation of the forward speed and heading. This paper extends this controller to automatically compute both of these inputs, in order to achieve trajectory planning in 3D cluttered environments. To do so, we first develop a method based on internal models, a concept widespread in cognitive neuroscience. We then compare the obtained gait to results generated with a more traditional planning method based on potential fields. In particular, we show that internal models result in more robust gaits by taking the walker dynamics into account.

Deploying humanoid robots in complex and unstructured environments requires the development of efficient and adaptive locomotion controllers. Bio-inspiration holds promises in this perspective, since humans are known to have both an energy efficient gait, and the capacity to modulate it across several features like forward speed and step length and height. In this paper, we report the development of a bio-inspired controller for bipedal walking that can achieve controlled modulations of the step height and length over a large range. This controller builds upon our previous work where we combined both a Central Pattern Generator (CPG) and reflex-like modulations with a layer of virtual muscles providing human-like leg impedance. Here, we report first a sensitivity analysis that was performed to identify those among the many parameters of our controller that can actually modulate the step height and length. Then, we report experimental results illustrating such controlled modulations over a large parameter space.

Reproducing human locomotion in simulation has a variety of applications, from informing prosthetic and rehabilitation medicine to generating stable and human-like robot or animated character movement. Previously, however, heavy optimizations were required to realize gaits specific to a single speed. Novel neuromuscular controllers blending feedforward and reflex-like control have shown state-of-the-art success in introducing speed modulation of walking gaits from a single optimization. In this work, we present a modified neuromuscular gait controller in the sagittal plane to similarly realize speed modulation for running gaits. As a result, our controller interpolates fewer than 10 parameters from a single optimization to realize a large range of running speeds on a simulated bipedal platform. We discuss the speed-evolution of these selected parameters, their significance in relation to human kinematics, and the velocity-tracking performance over the speed range between 1.3 m/s and 1.7 m/s, which scales to cover much of human running speeds.

Humanoid robots are gaining much interest nowadays. This is partly motivated by the ability of such robots to replace humans in dangerous environments being specifically designed for humans, such as man-made or natural disaster scenarios. However, existing robots are far from reaching human skills regarding the robustness to external perturbations required for such tasks, although torque-controlled and even bio-inspired robots hold new promises for research. A humanoid robot robustly interacting with its environment should be capable of handling highly uncertain ground structures, collisions, and other external perturbations. In this paper, a 3D bio-inspired balance controller is developed using a virtual lower limbs musculoskeletal model. An inverse muscular model that transforms the desired torque patterns into muscular stimulations closes the gap between traditional and bio-inspired controllers. The main contribution consists in developing a neural controller that computes the muscular stimulations driving this musculoskeletal model. This neural controller exploits the inverse model output to progressively learn the appropriate muscular stimulations for rejecting disturbances, without relying on the inverse model anymore. Two concurrent approaches are implemented to perform this autonomous learning: a cerebellar model and a support vector regression algorithm. The developed methods are tested in the Robotran simulation environment with COMAN, a compliant child-sized humanoid robot. Results illustrate that - at the end of the learning phase - the robot manages to reject perturbations by performing a full-body compensation requiring neither to solve an inverse dynamic model nor to get force measurement. Muscular stimulations are directly generated based on the previously learned perturbations.

The human foot plays a key role in human walking providing, among others, body support and propulsion, stability of the movement and impact absorption. These fundamental functionalities are accomplished by an extraordinarily rich bio-mechanical design. Nonetheless, humanoid robots follow different approaches to walk, hence, they generally implement rigid feet. In this study, we target the gap existing between the human foot and traditional humanoid-robot feet. More specifically, we evaluate the resulting advantages and draw-backs by implementing on a humanoid robot some of the properties and functionalities embedded in the human foot. To this end, we extract the physical characteristics of a prosthetic foot to develop a human-like foot model. This foot model is systematically tested in simulation in human-like walking tasks on flat ground and on uneven terrain. The movement of the limbs is generated by a muscle-reflex controller based on a simplified model of the human limbs. The gait features and the walking stability are evaluated for the human-like foot and compared with the results produced using rigid feet.

Bipedal walking with humanoid robots requires efficient real-time control. Nowadays, most bipedal robots require to ensure local stability at every instant in time, preventing them from achieving the impressive human walking skills. At the same time, bio-inspired walking controllers are emerging, though they are still mostly explored in simulation studies. However, porting these controllers to real hardware is needed to validate their use on real robots, as well as adapting them to face the world non-idealities. Here, we implemented one of them on a real humanoid robot, namely the COMAN, by conducting dynamic walking experiments. More precisely, we used a muscle-reflex model producing efficient and humanlike gaits. Starting from an off-line optimization performed in simulation, we present the controller implementation, focussing on the additional steps required to port it to real hardware. In our experimental results, we highlight some discrepancies between simulation and reality, together with possible controller extensions to fix them. Despite these differences, the real robot still managed to perform dynamic walking. On top of that, its gait exhibited stretched legs and foot roll at some points of the gait, two human walking features hard to achieve with most robot gaits. We present this on a 50 steps walk where the robot was free to move in the sagittal plane while lateral balance was provided by a human operator.

In this paper, we discuss some very important features for getting exploitable simulation results for multibody systems, relying on the example of a humanoid robot. First, we provide a comparison of simulation speed and accuracy for kinematics modeling relying on relative vs. absolute coor- dinates. This choice is particularly critical for mechanisms with long serial chains (e.g. legs and arms). Compliance in the robot actuation chain is also critical to enhance the robot safety and en- ergy efficiency, but makes the simulator more sensitive to modeling errors. Therefore, our second contribution is to derive the full electro-mechanical model of the inner dynamics of the compliant actuators embedded in our robot. Finally, we report our reasoning for choosing an appropriate contact library. The recommended solution is to couple our simulator with an open-source contact library offering both accurate and fast full-body contact modeling.

Controllers based on neuromuscular models hold the promise of energy-efficient and human-like walkers. However, most of them rely on optimizations or cumbersome hand-tuning to find controller parameters which, in turn, are usually working for a specific gait or forward speed only. Consequently, designing neuromuscular controllers for a large variety of gaits is usually challenging and highly sensitive. In this contribution, we propose a neuromuscular controller combining reflexes and a central pattern generator able to generate gaits across a large range of speeds, within a single optimization. Applying this controller to the model of COMAN, a 95 cm tall humanoid robot, we were able to get energy-efficient gaits ranging from 0.4 m/s to 0.9 m/s. This covers normal human walking speeds once scaled to the robot height. In the proposed controller, the robot speed could be continuously commanded within this range by changing three high-level parameters as linear functions of the target speed. This allowed large speed transitions with no additional tuning. By combining reflexes and a central pattern generator, this approach can also predict when the next strike will occur and modulate the step length to step over a hole.

Humanoid robots are currently still far from reaching the impressive human walking capabilities. Among the different methods used to design walking controllers, those based on the Zero-Moment Point (ZMP) criterion are among the most popular, even if they induce intrinsic limitations in terms of energy consumption and robustness. In parallel, bio-inspired controllers are emerging. They overcome the ZMP-based limitations, but still miss robust stabilization rules to be validated on real robots. This contribution studies how to efficiently compute the ZMP in realtime on a robot walking with bio-inspired control rules, in order to detect when the robot stability is compromised.