Dynamic Digital Human Model control design for the assessment of the workstation ergonomics ; Étude et conception de la commande de mannequins virtuels dynamiques pour l'évaluation ergonomique des postes de travail

The present thesis is focused on the evaluation of the biomechanical risk factors of a work-station or work-situation at the design stage. Generally, this evaluation is based upon experimental observations on comparable work stations or physical prototypes.The processed data in this case are mainly obtained from questionnaires, video analysis and interviews. The collection of precise quantitative data about postures and efforts requires complex and expensive technologies : electromyography, force sensors, motion capture system, etc. The use of those techniques can even hamper the execution of the analyzed task. An alternative solution consists in running simulations directly based on the future workstation digital design. Digital human models (DHM) with such features are currently available in computer-aided design (CAD) software.However, those digital human models present average characteristics and rely on static models, which induce in certain cases an underestimation of the operator's constraints. Consequently, an ambitious objective consists in developing a virtual human able to automatically calculate, from a simple description of the task, realistic motion in terms of position, speed, acceleration and applied couples, in order to obtain reliable ergonomic evaluation of a workstation. To reach this objective, we have developed a dynamic digital human model controlled by force and acceleration, inspired by human motor control. Our controller executes several tasks simultaneously in real time. Control algorithms are based upon the human ability to adapt and interact with different environments by adjusting force and impedance. The controller combines a multi-objective optimization problem (managing balance and contacts, manipulation) with a human learning model (humans would learn new dynamics because of a minimization of instability, error and effort). Theinterest of this method has been underlined through different simulations.This validation is based upon comparison of experimentally-measured ...