The main objective of this thesis is to propose new methods for the calibration of a large scale cable-driven robot. The principal method to improve the global behavior of a robot consists to identify the parameters of the model. For this, it is important to get redundant information by measuring the state of the robot in different configurations. However, the model used is a compromise between its ability to represent the actual behavior of the manipulator and the information available to fill in it. In the special case of the large scale cable-driven robots, mass and elasticity of the cables have a significant influence on the behavior of the robot but they are difficult to model. Indeed, the physical model of the cable is complex and requires knowledge of the tension inside it. Available sensors cannot provide this information with a sufficient accuracy to fill in a model of a realistic cable, we thus propose to use a simplified model. In order to provide an efficient calibration, it is necessary to fix the requirements to use this simplified model. Then, we have adapted and implemented some classical techniques for the calibration of parallel robots, but we also developed more innovative approaches. We propose a model for cable robots based on a representation of the uncertainties from modeling, measurements and parameters using intervals. By exploiting the interval analysis, we have developed various approaches to identify with certification the geometric parameters of the structure. We thus propose a new approach and associated algorithms to characterize and compute different kind of solutions for the calibration problem.
