Consistent kinematics and dynamics calibration of lightweight redundant industrial manipulators

Abstract

Absolute accuracy is one of industrial manipulator’s key performance characteristics, which is critical for emerging robotics applications such as laser cutting, riveting, and carbon fiber placement as well as for many machining operations. On the other hand, arrival of new uses such as collaborative robots needs the estimation of interaction efforts with the operator or with the environment (hand-guiding, collision detection, and free backlash assembly). This paper presents an approach to organize an integrated kinematic and dynamic calibration procedure to improve quality of models appropriate for trajectory planning and motion control. Along with bringing theoretical insights and novel arguments, we give hands-on recommendations on selection of parameters priors, initial guesses on calibration poses and trajectories, setting active constraints, algorithms tuning, and experimental data filtering which is necessary to perform consistent robot calibration in practice. We illustrate the study with experimental data and description of actual calibration performed on the KUKA Light-Weight Robot using vision-based metrology and dedicated software. In contrast to authors preceding works, this paper includes a more complete entire procedure description, analysis of dynamic calibration sensitivity with respect to kinematic parameters estimates and a chapter on how calibration results can be used for model-based trajectories planning using virtual holonomic constraints approach.