Skip to Main content Skip to Navigation

Estimation and control of the deformations of an exoskeleton using inertial sensors

Abstract : This thesis addresses the general problem of the compensation of deformations on a walking robot, and considers it in the particular context of the medical exoskeleton Atalante. Structural deformations are unavoidable in anthropomorphic robots with long legs and a heavy torso. This is even more true for an exoskeleton, which has to support the weight of the user. Meanwhile, the presence of an uninstrumented human user leads to significant disturbances, that limit the performance of feedforward corrections. We propose a sensor-based methodology, that feedbacks the data of several Inertial Measurement Units (IMUs) onto the actuators, in order to mitigate the effect of the multiple flexibilities on Atalante. To that end, we kinematically model the deformations as extra spherical joints. We study several observers to estimate the rotations induced by the flexibilities, by estimating the attitude of the IMUs. Three observers, based on various models, are considered: a (naive) zero-on-average acceleration model, a kinematic model and a dynamic model. We conclude that the best results are obtained by relying only on the kinematic model of the robot to build an approximate velocity measurement. This allows the design of an observer able to handle strong accelerations while being robust to the dynamical uncertainties linked to patient behavior. These attitude estimates are then projected onto the joints of the robot, to adopt a joint elasticity model, which we use to perform decentralized high-gain feedback control. This methodology is experimentally validated on Atalante, where it yields improved disturbance rejection and improved trajectory tracking, enhancing the robustness of Atalante's walk with a user.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, February 7, 2022 - 8:37:08 AM
Last modification on : Wednesday, February 9, 2022 - 3:06:00 AM
Long-term archiving on: : Sunday, May 8, 2022 - 6:14:28 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03559553, version 1


Matthieu Vigne. Estimation and control of the deformations of an exoskeleton using inertial sensors. Robotics [cs.RO]. Université Paris sciences et lettres, 2021. English. ⟨NNT : 2021UPSLM044⟩. ⟨tel-03559553⟩



Record views


Files downloads