Abstract
The work of this thesis contributes to solve one of the actual issues of research on parallel robots which is<br />to find a compromise between high dynamics and good accuracy. Three approaches have been considered :<br />geometrical calibration using external measurements, use of metrological redundancy, and finally application<br />of the dissociated measurement principle. First, a classical geometrical calibration of a linear Delta<br />robot and the first prototype of the Par4 robot are presented. Despite being effective in the case of quasistatic<br />problems, this method reveals to be insufficient when the forces applied to the robot's structure bring<br />about deformations on the latter. Parallel robots with actuation redundancy thus present a particular difficulty<br />to be calibrated since internal constraints may appear in their mechanisms and deform their structure.<br />This is why a method to transform actuation redundancy into kinematic redundancy has been proposed.<br />Metrological redundancy has also been studied, and several methods benefitting from it have been<br />analyzed and tested. Finally, the concept of dissociation between actuation and measurement is introduced.<br />This novel concept in parallel robotics lies on the separation of motion and force transmission from the<br />robot's state measurement. This concept has been applied on a parallel-based machine tool, and a prototype<br />has been built. Control laws have been tested, and results concerning precision improvement are presented