Commande basée intention d'un exosquelette des membres inférieurs.

par José Colorado (IVáN)

Projet de thèse en Signal, Image, Automatique

Sous la direction de Yacine Amirat.

Thèses en préparation à Paris Est , dans le cadre de MSTIC : Mathématiques et Sciences et Technologies de l'Information et de la Communication , en partenariat avec LISSI - Laboratoire Images, Signaux et Systèmes Intelligents (laboratoire) depuis le 01-09-2015 .

  • Titre traduit

    Intention-based control of a lower limb exoskeleton.


  • Résumé

    Assistance/rehabilitation robot applications based on human-robot interactions have gained increasing interest over the past decade with the continuous rise of the aged and dependent population. Nowadays, the rise in life expectancy is set to continue, combined with falling birth rates, will certainly accelerate the ageing of the population. Dependent people (patients and elderly) have limited capabilities to achieve various tasks of the daily life activities. This issue can be faced using a program of rehabilitation exercises. The use of exoskeletons can offer a solution for the rehabilitation tasks that are often repetitive and require some endurance from both therapist and patient. Exoskeletons are wearable mechanical external devices with actuated joints associated to those of the human body. One can notice two major challenges when using exoskeletons: the first one is related to the mechanical aspect, and consists in the design of an appropriate mechanical structure that offers good ergonomics and optimal energy transfer to the wearer. The second challenge regards the interaction between the exoskeleton and the wearer. Physical and cognitive interactions between the wearer and the exoskeleton are explored to have an optimal use of the exoskeletons. Cognitive interaction can be used to adapt the assisting torque generated by the exoskeleton upon the wearer intention initiated by the Central Nervous System (CNS). Two approaches of rehabilitation/assistance will be studied in this thesis. The first one deals with the passive rehabilitation where no effort is delivered by the wearer. It consists of performing repetitive motions of the limbs following a desired trajectory predetermined by a rehabilitation therapeutic doctor. This approach concerns mainly patients having lost total control of their lower limbs due to spinal cord injuries. Within the second approach, a human-exoskeleton interaction strategy is proposed to provide torque assistance upon the wearer intention in order to produce voluntary lower limb movements. The goal of this thesis proposal is the development of wearable robot control techniques to assist people suffering from lower limb muscle impairments during knee joint flexion/extension, sit-to-stand, stand- to-sit actions and walking movements. More precisely, the study aims to determine a precise kinematic and dynamic model of the lower limbs during daily living activities exerted by dependent people in a free and unconstrained environment. The wearable robot should be actuated in response to the wearer's intention, computed through the movement kinematics of the wearer's lower limbs. Detection algorithms and observers should be sufficiently robust to detect and filter undesirable signals such as sudden movements (muscular spasm) of the wearer. Non-linear control laws will be particularly developed and should be computed according to the wearer's intention to assist his/her lower limbs movements. These control laws must be sufficiently robust with respect to modeling or identification errors as well as sensor noises. Regarding the estimation of lower limbs motion intention, the human-robot interface is commonly modeled using wearable sensors, such as electromyography (EMG), Ground Reaction Forces (GRF) and torque sensors. Since the EMG signals are able to predict the human motion intention directly, they are widely used for the control of exoskeletons. The use of the aforementioned sensors for estimating the human motion intention is usually facing several practical problems. The EMG sensors are sensitive to many factors, such as electrodes placement, neighboring muscles signals, and noises that affect the original EMG recordings, etc. In order to accurately estimate the human joint torque by using GRF, the tangential force between the shoes and the foot is required, which is usually difficult to be measured. In terms of use of torque/force sensors, the applicable places are constrained for many compact exoskeletons. In this thesis proposal, a new paradigm related to the intention-based nonlinear impedance control structure for the lower limb exoskeleton, will be developed. The goal of this control structure is to decrease the impedance of the human-exoskeleton system to a desired level. Since the desired impedance is smaller than the one of the human lower limb, the human muscle torque or metabolic cost of the wearer can be decreased as well. A nonlinear observer will be used to estimate the voluntary human torque generated by the wearer. A desired impedance model we be designed to estimate the motion of the human lower limbs. In order to take into account the parametric uncertainties, a Nonlinear Disturbance Observer (NDO) will be used to ensure satisfactory tracking of the estimated lower limb human joint movements. The thesis will include experimental validation that will be done in clinical settings in close collaboration with the CHU Henry Mondor, which is located nearby the University of Paris-Est Créteil (UPEC).