19documents trouvés
Reverse control for humanoid robot task recognition
S.HAK, N.MANSARD, O.STASSE, J.P.LAUMOND
GEPETTO
Revue Scientifique : IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, Vol.42, N°6, pp.1524-1537, Décembre 2012, DOI 10.1109/TSMCB.2012.2193614 , N° 12257
Lien : http://hal.archives-ouvertes.fr/hal-00697272
Diffusable
Plus d'informations
Abstract
Efficient methods to perform motion recognition have been developed using statistical tools. Those methods rely on primitives learning in a suitable space, for example, the latent space of the joint-angle and/or adequate task spaces. Learned primitives are often sequential : a motion is segmented according to the time axis. When working with a humanoid robot, a motion can be decomposed into parallel sub-tasks. For example, in a waiter scenario, the robot has to keep some plates horizontal with one of its arms, while placing a plate on the table with its free hand. Recognition can thus not be limited to one task per consecutive segment of time. The method presented in this paper takes advantage of the knowledge of what tasks the robot is able to do and how the motion is generated from this set of known controllers, to perform a reverse engineering of an observed motion. This analysis is intended to recognize parallel tasks that have been used to generate a motion. The method relies on the task-function formalism and the projection operation into the null space of a task to decouple the controllers. The approach is successfully applied on a real robot to disambiguate motion in different scenarios where two motions look similar but have different purposes.
Walking on non-planar surfaces using an inverse dynamic stack of tasks
O.RAMOS PONCE, N.MANSARD, O.STASSE, P.SOUERES
GEPETTO
Manifestation avec acte : IEEE-RAS International Conference on Humanoid Robots ( HUMANOIDS ) 2012 du 29 novembre au 01 décembre 2012, Osaka (Japon), 2012, 7p. , N° 12654
Lien : http://hal.archives-ouvertes.fr/hal-00738243
Diffusable
Plus d'informations
Abstract
This paper presents a method to handle walking on non-planar surfaces. The method obtains the trajectory of the center of mass and the next position of the foot from a pattern generator. Then, an inverse dynamics control scheme with a quadratic programming optimization solver is used to let the foot go from its initial to its final position, controlling also the center of mass and the waist. This solver is able to handle an arbitrary number of contact points. When the swinging foot is going down, collision points are detected and they are added as contact points to the model as soon as they appear. If there are three or more contact points, the foot can safely step, but if there are one or two contact points, the foot rotates properly to generate the largest support polygon. Using this heuristic, the foot can stand on non-planar surfaces. The results show the simulation of HRP-2 walking on a surface with obstacles.
Reliable indoor navigation on humanoid robots using vision-based localization
T.MOULARD, P.FERNANDEZ ALCANTARILLA, F.LAMIRAUX, O.STASSE, F.DELLAERT
GEPETTO, ISIT, Georgia Institute
Rapport LAAS N°12549, Octobre 2012, 7p.
Lien : http://hal.archives-ouvertes.fr/hal-00733666
Diffusable
Plus d'informations
Abstract
Reliable localization on a humanoid robot is a sine qua non condition to succeed in realizing complex robotics scenarios. Before dealing with perturbations and online modification of the environment, one has to make sure that the planned trajectory alone will be correctly followed. This paper demonstrates that a control framework suited to humanoid robots relying on a vision-based localization system can achieve this goal. Our localization framework is based on a real-time vision-based localization system that assumes that a pre-existing 3D map of the environment exists and allows to obtain accurate results in complex robotics scenarios. By compensating for execution errors such as drifts and robot model errors, the HRP-2 robot is able to achieve high precision tasks.
Trajectory following for legged robots
T.MOULARD, F.LAMIRAUX, O.STASSE
GEPETTO, AIST
Manifestation avec acte : International Conference on Biomedical Robotics and Biomechatronics (BioRob'2012), Rome (Italie), 24-28 Juin 2012, 6p. , N° 11429
Lien : http://hal.archives-ouvertes.fr/hal-00601291/fr/
Diffusable
Plus d'informations
Abstract
While robust trajectory following is a well-studied problem on mobile robots, the question of how to track accurately a trajectory on a humanoid robot remains open. This paper suggests a closed-loop trajectory tracking strategy aimed at humanoid robots. Compared to approaches from mobile robotics, this control scheme takes into account footsteps alteration, equilibrium constraints and singularities avoidance for humanoids. It provides a robust way to execute long and/or precise motion with the ability of correcting on-line preplanned trajectories in a very reactive manner. Results have been validated on the HRP-2 humanoid platform.
Real-time footstep planning for humanoid robots among 3D obstacles using a hybrid bounding box
N.PERRIN, O.STASSE, F.LAMIRAUX, Y.J.KIM, D.MANOCHA
GEPETTO, AIST, Seoul, North Carolina
Manifestation avec acte : IEEE International Conference on Robotics and Automation (ICRA 2012), St Paul (USA), 14-18 Mai 2012, pp.977-982 , N° 12055
Lien : http://hal.archives-ouvertes.fr/hal-00668794
Diffusable
Plus d'informations
Abstract
In this paper we introduce a new bounding box method for footstep planning for humanoid robots. Similar to the classic bounding box method (which uses a single rectangular box to encompass the robot) it is computationally efficient, easy to implement and can be combined with any rigid body motion planning library. However, unlike the classic bounding box method, our method takes into account the stepping over capabilities of the robot, and generates precise leg trajectories to avoid obstacles on the ground. We demonstrate that this method is well suited for footstep planning in cluttered environments.
Capture, recognition and imitation of anthropomorphic motion
S.HAK, N.MANSARD, O.RAMOS PONCE, L.SAAB, O.STASSE
GEPETTO
Manifestation avec acte : IEEE International Conference on Robotics and Automation (ICRA 2012), St Paul (USA), 14-18 Mai 2012, pp.3539-3540 , N° 12276
Lien : http://hal.archives-ouvertes.fr/hal-00706661
Diffusable
Plus d'informations
Abstract
This work presents an overview of our current research works in generation, recognition and editing of anthropomorphic motion using a unified framework: the stack of tasks. It is based on the task function formalism classically used for motion generation. A task function maps the joint space of a robot to a dedicated space which is usually linked to the sensors of the robot: the task space. The task spaces are suitable to perform motion analysis and task recognition because the tasks are described in those spaces. The generation is originally based on inverse kinematics but can be generalized to produce full-dynamic motions. The tasks are defined by a tasks space, a reference behavior and a task Jacobian. The reference behaviors are originated from human trajectories. Specific tasks are then integrated to retarget and to edit the reference motion in order to respect the dynamic constraints, the limits of the robot and the general aspect.
Fast humanoid robot collision-free footstep planning using swept volume approximations
N.PERRIN, O.STASSE, L.BAUDOUIN, F.LAMIRAUX, E.YOSHIDA
AIST, GEPETTO
Revue Scientifique : IEEE Transactions on Robotics, Vol.28, N°2, pp.427-439, Avril 2012 , N° 12166
Lien : http://hal.archives-ouvertes.fr/hal-00662550
Diffusable
Plus d'informations
Abstract
In this paper, we propose a novel and coherent framework for fast footstep planning for legged robots on a flat ground with 3D obstacle avoidance. We use swept volume approximations computed offline in order to considerably reduce the time spent in collision checking during the online planning phase, in which an RRT variant is used to find collision-free sequences of half-steps (produced by a specific walking pattern generator). Then, an original homotopy is used to smooth the sequences into natural motions avoiding gently the obstacles. The results are experimentally validated on the robot HRP-2.
Real-time replanning using 3D environment for humanoid robot
L.BAUDOUIN, N.PERRIN, T.MOULARD, F.LAMIRAUX, O.STASSE, E.YOSHIDA
GEPETTO, AIST
Manifestation avec acte : IEEE-RAS International Conference on Humanoid Robots (HUMANOIDS 2011), Bled (Slovénie), 26-28 Octobre 2011, pp.584-589 , N° 11369
Lien : http://hal.archives-ouvertes.fr/hal-00601300/fr/
Diffusable
Plus d'informations
Abstract
In this paper, we illustrate experimentally an origi- nal real-time replanning scheme and architecture for humanoid robot reactive walking. Based on a dense set of actions, our approach uses a large panel of the humanoid robot capabilities and is particularly well suited for 3D collision avoidance. Indeed A-* approaches becomes difficult in such situation, thus the method demonstrated here relies on RRT. Combined with an approximation of the volume swept by the robot legs while walking, our method is able to cope with 3D obstacles while maintaining real-time computation. We experimentally validate our approach on the robot HRP-2.
Weakly collision-free paths for continuous humanoid footstep planning
N.PERRIN, O.STASSE, F.LAMIRAUX, E.YOSHIDA
GEPETTO, AIST
Manifestation avec acte : IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2011), San Francisco (USA), 25-30 Septembre 2011, pp.4408-4413 , N° 11577
Lien : http://hal.archives-ouvertes.fr/hal-00639026/fr/
Diffusable
Plus d'informations
Abstract
In this paper we demonstrate an original equivalence between footstep planning problems, where discrete sequences of steps are searched for, and the more classical problem of motion planning for a 2D rigid shape, where a continuous collision-free path has to be found. This equivalence enables a lot of classical motion planning techniques (such as PRM, RRT, etc.) to be applied almost effortlessly to the specific problem of footstep planning for a humanoid robot.
Adaptive sampling-based approximation of the sign of multivariate real-valued functions
N.PERRIN, O.STASSE, F.LAMIRAUX, E.YOSHIDA
GEPETTO, AIST
Rapport LAAS N°11229, Mai 2011, 5p.
Lien : http://hal.archives-ouvertes.fr/hal-00544891/fr/
Diffusable
Plus d'informations
Abstract
In this technical report, we present an algorithm which approximates the sign of multivariate real-valued functions through adaptive sampling, and demonstrate a convergence result. The algorithm was specifically designed for real-time robotics applications where quick evaluations are often needed.