Publications

Filter by type:

Experimental Evaluation of Simple Estimators for Humanoid Robots. . Humanoids, 2017.

PDF

This paper introduces and evaluates a family of new simple estimators to reconstruct the pose and velocity of the floating base. The estimation of the floating-base state is a critical challenge to whole-body control methods that rely on full-state information in high-rate feedback. Although the kinematics of grounded limbs may be used to estimate the pose and velocity of the body, modelling errors from ground irregularity, foot slip, and structural flexibilities limit the utility of estimation from kinematics alone. These difficulties have motivated the development of sensor fusion methods to augment body-mounted IMUs with kinematic measurements. Existing methods often rely on extended Kalman filtering, which lack convergence guarantees and may present difficulties in tuning. This paper proposes two new simplifications to the floating-base state estimation problem that make use of robust off-the-shelf orientation estimators to bootstrap development. Experiments for in-place balance and walking with the HRP-2 show that the simplifications yield results on par with the accuracy reported in the literature for other methods. As further benefits, the structure of the proposed estimators prevents divergence of the estimates, simplifies tuning, and admits efficient computation. These benefits are envisioned to help accelerate the development of baseline estimators in future humanoids.
Multi-contact Locomotion of Legged Robots in Complex Environments – The Loco3D project. . RSS Workshop, 2017.

PDF

Planning, adapting and executing multi-contact locomotion movements on legged robots in complex environments remains an open problem. In this proposal, we introduce a complete pipeline to address this issue in the context of humanoid robots inside industrial environments. This pipeline relies on a multi-stage approach in order to simplify the process flow and to exploit at best state-of-the-art techniques both in terms of contact planning, whole-body control and perception. The main challenges lie in the choice of the different modules composing this pipeline as well as their mutual interactions: e.g. at which frequency rates each module has to work in order to allow safe and robust locomotion? or which information must transit between the modules? We named this project Loco3D standing for Locomotion in 3D, in contrast to the classic locomotion on quasi-flat terrains, where the motion of the center of mass of the robot is mostly limited to a 2D plane.
Estimation stabilisation de l'état d'un robot humanoïde compliant. . INP Toulouse (France), 2017.

PDF Slides

Cette thèse traite de l’estimation et de la stabilisation de l’état des compliances passives présentes dans les chevilles du robot humanoïde HRP-2. Ces compliances peuvent être vues comme un degré de liberté unique et observable, sous quelques hypothèses qui sont explicitées. L’estimateur utilise des mesures provenant de la centrale inertielle située dans le torse du robot et éventuellement des capteurs de forces situés dans ses pieds. Un filtre de Kalman étendu est utilisé pour l’estimation d’état. Ce filtre utilise un modèle complet de la dynamique du robot, pour lequel la dynamique interne du robot, considérée comme parfaitement connue et contrôlée, a été découplée de la dynamique de la compliance passive du robot. L’observabilité locale de l’état a été montrée en considérant ce modèle et les mesures provenant de la centrale inertielle seule. Il a de plus été montré que l’ajout des mesures des capteurs de forces dans les pieds du robot permet de compléter l’état avec des mesures d’erreurs dans le modèle dynamique du robot. L’estimateur a été validé expérimentalement sur le robot humanoïde HRP-2. Sur cet estimateur a été construit un stabilisateur de l’état de la compliance d’HRP-2. L’état commandé est la position et vitesse du centre de masse (contrôle indirecte de la quantité de mouvement) du robot, l’orientation et la vitesse angulaire de son tronc (contrôle indirecte du moment cinétique), ainsi que l’orientation et la vitesse angulaire de la compliance. Les grandeurs de commande sont l’accélération du centre de masse du robot et l’accélération angulaire de son tronc. Un régulateur quadratique linéaire (LQR) a été utilisé pour calculer les gains du retour d’état, basé sur un modèle appelé “pendule inverse flexible à roue d’inertie” qui consiste en un pendule inverse dont la base est flexible et où une répartition de masse en rotation autour du centre de masse du robot représente le tronc du robot. Des tests ont été effectués sur le robot HRP-2 en double support, utilisant l’estimateur décrit précédemment avec ou sans les capteurs de forces.
Stabilization of a Compliant Humanoid Robot Using Only Inertial Measurement Units with a Viscoelastic Reaction Mass Pendulum Model. . IROS, 2016.

PDF Video

To guarantee its balance, a humanoid robot has to respect some contact force constraints. Therefore, traditional controllers generate motions complying with these constraints, but they usually consider the robot as stiff and the joint position perfectly known. However, several robots contain compliant parts in their structure. This flexibility modifies the forces at contacts and endangers balance. However, most solutions to stabilize the robot rely on force sensors. But several humanoid robots aren’t equipped with these sensors. This paper has two aims. The first one is to develop a compliance stabilizer using the center of mass position and upper-body orientation through a viscoelastic reaction mass pendulum model. The second objective is to show the performances of such a stabilizer when relying only on an IMU-based state observer. Experimental results on HRP-2 robot show that the stabilization successfully rejects perturbations with high gains using only these IMU signals. Moreover, the actuation of the upper-body orientation provides redundancy, robustness and finally improved performances to the stabilizer.
Fusion of Force-Torque Sensors, Inertial Measurements Units and Proprioception for a Humanoid Kinematics-Dynamics Observation. . Humanoids, 2015.

PDF

We present a scheme where the measurements obtained through inertial measurement units (IMU), contact-force sensors and proprioception (joint encoders) are merged in order to observe humanoid unactuated floating-base dynamics. The sensor data fusion is implemented using an Extended Kalman Filter. The prediction part is constituted by viscoelastic contacts assumption and a model expressing at the origin the full body dynamics. The correction is achieved using embedded IMU and force sensor. Simulation and experimentation on HRP-2 robot show a state observation with improves inter-sensor consistency but also increased reconstruction accuracy.
Estimation of Contact Forces and Floating Base Kinematics of a Humanoid Robot Using Only Inertial Measurement Units. . IROS, 2015.

PDF Video

A humanoid robot is underactuated and only relies on contacts with environment to move in the space. The ability to measure contact forces and torques enables then to predict the robot dynamics including balance. In classical cases, a humanoid robot is considered as a multi-body system with rigid limbs and joints and interactions with the environment are modeled as stiff contacts. Forces and torques at contacts are generally estimated with sensors which are expensive and sensitive to calibration errors. However, a robot is not perfectly rigid and contacts may have flexibilities. Therefore, external forces create geometric deformations of the body or its environment. These deformations may modify the robot dynamics and produce unwanted and unbalanced motions. Nonetheless, if we have a model of contact stiffness and are able to reconstruct reliably the geometric deformation, we can reconstruct forces and torques at contact. This study aims at estimating contact forces and torques and to observe the body kinematics of the robot with only an Inertial Measurements Unit (IMU). We show that we are able to reconstruct efficiently the position of the Center of Pressure (CoP) of the robot with only the IMU and proprioceptive data from the robot.
Localization of multiple sources from a binaural head in a known noisy environment. . IROS, 2014.

PDF

This paper presents a strategy to the localization of multiple sound sources from a static binaural head. The sources are supposed W-Disjoint Orthogonal and their number is assumed known. Their most likely azimuths are computed by means of the Expectation-Maximization algorithm. Application of the method on simulated data is reported, as well as some evaluations of its HARK implementation on experimental data. Two important properties are observed: scattering effects can be coped with, thanks to the required prior knowledge of the (room-independent) head interaural transfer function; the environment noise statistics are handled separately.