Search

International audience; Cet article présente un ensemble d'outils génériques en dynamique des systèmes mul-ticorps consacrés à l'étude de la locomotion bio-inspirée. Partant d'exemples empruntés à la nature et la robotique, nous poserons un problème général de locomotion dont la résolution nous permettra d'installer progressivement un cadre géométrique unifié dédié à ce problème. Pour cela, nous partirons du modèle des systèmes multicorps mobiles dis-crets, que nous étendrons progressivement au cas des systèmes continus puis mous. Nous aborderons également le problème pratique de l'implémentation efficace de ces modèles en proposant une approche basée sur la méthode de Newton-Euler, illustrée par quelques exemples liés à la reptation, la natation et le vol. Abstract This article presents a set of generic tools of Multibody Systems dynamics devoted to the study of bioinspired locomotion in robotics. Starting from examples borrowed from nature and robotics, we will pose a general problem of locomotion whose resolution will allow us to progressively install a unified geometric framework dedicated to this problem. For that purpose, we start from the model of discrete mobile multibody systems, that we will progressively extend to the case of continuous and finally soft systems. Beyond these theoretical aspects, we address the practical problem of the efficient computation of these models by proposing an efficient locomotion dynamics Newton-Euler based approach with a few illustration on creeping, swimming and flying.

International audience; The paper deals with the dynamic modeling of bio-inspired robots with soft appendages such as flying insect-like or swimming fish-like robots. In order to model such soft systems, we propose to use the Mobile Multibody System framework introduced in [1][2][3]. In such a framework, the robot is considered as a tree-like structure of rigid bodies where the evolution of the position of the joints is governed by stress-strain laws or control torques. Based on the Newton-Euler formulation of these systems, we propose a new algorithm able to compute at each step of a time loop both the net and passive joint accelerations along with the control torques supplied by the motors. To illustrate, based on previous work [4], the proposed algorithm is applied to the simulation of the hovering flight of a soft flapping-wing insect-like robot (see the attached video).

International audience; This article presents a unified dynamic modeling approach of continuum robots. The robot is modeled as a geometrically exact beam continuously actuated through an active strain law. Once included into the geometric mechanics of locomotion, the approach applies to any hyper-redundant or continuous robot devoted to manipulation and/or locomotion. Furthermore, exploiting the nature of the resulting models as being a continuous version of the Newton-Euler models of discrete robots, an algorithm is proposed which is capable of computing the internal control torques (and/or forces) as well as the rigid overall motions of the locomotor robot. The efficiency of the approach is finally illustrated through many examples directly related to the terrestrial locomotion of elongated animals as snakes, worms or caterpillars and their associated bio-mimetic artifacts.

International audience; In this article, we propose a dynamic model of the three-dimensional eel swim. This model is analytical and suited to the on-line control of eel-like robots. The proposed solution is based on the Large Amplitude Elongated Body Theory of Lighthill and a working frame recently proposed in [1] for the dynamic modeling of hyper-redundant robots. This working frame was named "macro-continuous" since at this macroscopic scale, the robot (or the animal) is considered as a Cosserat beam internally (and continuously) actuated. This article proposes new results in two directions. Firstly, it achieves an extension of the Lighthill theory to the case of a self propelled body swimming in three dimensions, while including a model of the internal control torque. Secondly, this generalization of the Lighthill model is achieved due to a new set of equations which is also derived in this article. These equations generalize the Poincaré equations of a Cosserat beam to the case of an open system containing a fluid stratified around the slender beam.

International audience

Proceedings of CIFA2008, Bucarest Romania, september 2008; This paper proposes and compares two observers designed to calculate the location of an obstacle. The two methods are bio-inspired with a sense used by electric fishes of equatorial forests: the electrolocation. Firstly, this study presents the electrolocation and then develops two models of emitter-sensors inspired by the electrical sense. Secondly, the two models are used in different observers for detection and localisation of wall obstacles. The estimation methods are based on an Extended Kalman Filter algorithm. Observers are tested on simulations in order to assess their potentials and to analyze observability.

International audience; This article proposes a dynamic model of the swim of elongated ﰣshes suited to the on-line control of bio-mimetic eel-like robots. The approach is analytic and can be considered as an extension of the original reactive "Large-Elongated-Body-Theory" of Lighthill to the three dimensional self propulsion augmented of a resistive empirical model. While all the mathematical fundamentals are detailed in [1], this article essentially focuses on the numerical validation and calibration of the model and the study of swimming gaits. The proposed model is coupled to an algorithm allowing us to compute the motion of the ﰣsh head and the ﰣeld of internal control torque from the knowledge of the imposed internal strain ﰣelds. Based on the Newton-Euler formalism of robots dynamics, this algorithm works faster than real time. As far as precision is concerned, many tests obtained with several planar and three dimensional gaits are reported and compared (in the planar case) with a Navier-Stokes solver, devoted until today to the planar swim. The comparisons obtained are very encouraging since in all the cases we tested, the diﰢerences between our simpliﰣed and reference simulations do not exceed ten per cent.

International audience

The presented work in this thesis is devoted to the development of a dynamic model for the an- guilliform swimming suited to the real time control of the future biomimetic "Eel-like robot" of the ROBEA-CNRS project of the same name. The computation of the interactions between a deformable body and the fluid, in which it moves, is a complex problem requiring the integration of the Navier-Stokes equations and the non-linear dynamic equations of the body enduring finished transformations. Pursuing goals of control for robotic, the suggested solution is based on the fusion of two theories : the Slender Body Theory of the fluid mechanic and Cosserat Beam Theory of the solid mechanic. The first theory models the 3-D flow around fish by the stratification "slices by slices" of the 2-D flows and transverse with the principal axis of the animal's body. Basing on the second theory, the fish is modeled by a continuous assembly of rigid sections that represents the animal's vertebrae or the eel-like robot's rigid platforms. On the basis of this model, the purpose of the presented work is to establish the dynamics of the head and the vertebrae of fish in order to work out a numerical algorithm based on the "Newton-Euler formalism". Finally, the obtained simulator works in real time with a good level of precision (i.e. lower than 10%) compared with the numerical computations of the Navier-Stokes equations.; Le travail présenté dans ce manuscrit est consacré à l'élaboration d'un modèle dynamique de la nage pour la commande du futur "Robot Anguille" du projet ROBEA-CNRS du même nom. Dans l'absolu, le calcul des interactions entre un corps déformable et le fluide sur lequel il s'appuie pour se déplacer, est un problème complexe nécessitant l'intégration des équations de Navier-Stokes couplées aux équations non-linéaires de la dynamique du corps soumis à des transformations finies. Poursuivant des objectifs de commande pour la robotique, la solution proposée dans ce travail est basée sur la fusion de deux théories : celle du "corps mince" issue de la mécanique des fluides et celle des "poutres Cosserat" de la mécanique du solide. La première théorie permet de remplacer l'écoulement 3-D autour du poisson par la stratification "tranche par tranche" d'écoulements plans, transverses à l'axe principal du corps de l'animal. Quant à la seconde, elle assimile le poisson à l'assemblage continu de sections rigides modélisant ses vertèbres ou, dans un contexte plus technologique, les plate-formes parallèles de notre robot bio-mimétique. Sur la base de cette modélisation, le travail présenté a pour but d'établir les dynamiques de la tête et des vertèbres du poisson afin d'élaborer in fine un algorithme de simulation numérique basé sur le "formalisme de Newton-Euler" de la robotique, ici étendu aux robots locomoteurs continus. Finalement, le modèle élaboré réalise une généralisation du modèle de Lighthill au cas de la nage tridimensionnelle d'un corps élancé autopropulsé. Outre ce résultat purement analytique, le simulateur qui en résulte nous a permis de mettre au point des allures jamais étudiées jusqu'alors. Qui plus est, il tourne en "temps réel", tout en maintenant un bon niveau de précision (i.e. inférieur à 10%) comparé à la référence basée sur la résolution numérique des équations de Navier-Stokes.