Vers un contrôle d'assistance intuitif d'un exosquelette via les signaux électromyographiques

Exoskeletons are promising devices, particularly for the prevention of musculoskeletal disorders, post-stroke rehabilitation or compensating for motor loss after spinal cord injury. However, they need to be properly controlled before they can be used in industry, hospitals or homes.To achieve this, it is first necessary to establish what assistance needs to be provided during the user's movement. The aim of this thesis work was therefore first to develop and evaluate intention detection techniques based on electromyographic signals. Evaluation of their accuracy showed that the movement the user wishes to make and the force he wishes to apply to his environment can be estimated in real time from these signals.Once this assessment had been made, this work was then dedicated to the development of an assistance controller exploiting this estimation. Several controllers were proposed and tested on the functional task of moving a mass from one point to another. Variables linked to the mechanical energy and intuitiveness of the movement were evaluated, demonstrating the reduction in effort required by the assisted user, and the existence of a compromise between intuitiveness and the strength of the assistance. This work opens up interesting prospects for the use of electromyographic signals to assist movement in an exoskeleton.