Browsing by Author "Schmidt, Valentin Lorenz"

Now showing 1 - 1 of 1

Open Access
Modeling techniques and reliable real-time implementation of kinematics for cable-driven parallel robots using polymer fiber cables
(Stuttgart : Fraunhofer Verlag, 2017) Schmidt, Valentin Lorenz; Pott, Andreas (Jun.-Prof. Dr.-Ing.)
In this thesis, the accuracy of cable-driven parallel robots is investigated by presenting and discussing relevant factors. Models to compensate factors affecting accuracy, including cable properties such as mass, elasticity, creep and ovalization, cable pulleys, and drive trains, are given and applied where necessary. Ovalization was previously ignored in literature, but it is shown to have a significant effect. Methods of estimating the impact of each factor are used to determine a classification for cable-driven parallel robots constructed with plastic fibers. Estimations indicate that some of the factors, such as pulleys, elongation and ovalization are more significant than cable mass, temperature and a non-linear transmission ratio. These findings give an indication of which factors should be modeled first. A focus lies in the real-time capability of the presented models. The incorporation of accuracy factors in the robot controller to improve the robot kinematics is not trivial, particularly for the forward kinematics. Methods for the numerical evaluation of the forward kinematics are thus presented. The most effective improvement is an adapted intersection method for estimating the position from given cable lengths. The intersection method works remarkably well for most geometry types. Further, it is shown that the type of numerical algorithm and value preconditioning affect the proficiency of numerical solvers. The models for improving accuracy are grouped into four architecture types: direct implementation, compensation, offline calculation, and sensor feedback. The direct architecture enables complex control algorithms but is only suitable for a few mathematical models. Compensation is applicable for a wide range of models and has the advantage of retaining reliability. Factors which cannot be compensated in real-time can also be calculated offline, and any factors which have additional measurable parameters need to be incorporated into a feedback control. However, cable forces can also be approximated to achieve a simple elongation compensation. In a practical investigation, the extended models which compensate factors affecting accuracy, are verified for two cable robots. Positional accuracy, positional repeatability, pose drift, posing time and static compliance are tested. For cable robots driven by plastic fiber cables, accuracy scales with size and is 1.38mm and 40.42mm for a robot with a 1.54m and 20.2m diagonal size respectively. The repeatability of the same robots is 0.0806mm and 5.24mm. There is a significant negative correlation between static compliance and accuracy. Improvements through applying extended models are verified. Positional accuracy is improved by 30% when using a simple elongation compensation in the case of the IPAnema 3 cable robot.