| dc.description.abstract | The paper introduces an enhanced calibration algorithm
that is robust and reliable for the kinematic calibration of robots and other mechanisms featuring redundant parameterization, unmeasured time-varying coordinates, kinematic singularities, and redundant kinematic equations. The algorithm is presented in general and then tested on two complex mechanisms. The initial test is a simulated calibration of a two-degree-of-freedom planar mechanism with five loops and eleven moving bodies. Subsequently, a 3[RRPaR] Delta robot is calibrated experimentally. As a reference, a standard calibration was realized with an ad hoc simplified model. Then, a fully general model was assembled and calibrated using the new method. Both calibrated models were integrated into the robot’s control system, and their positioning accuracy was assessed on a 160-point trajectory spanning the workspace. The new method’s positioning error after calibration was, on average, 25% lower compared to the ad hoc model approach. The new algorithm does not require ad hoc search for formulation, parameterization, or singularity analysis. The model can be built automatically and straightforwardly without the danger of potential under-parameterization and, thus, a reduction in positioning accuracy. | cze |
