Traceable digital twin for accurate positioning of industrial robot arms in human–robot collaborative systems
Abstract
Human–robot collaborative (HRC) systems are getting increasing attention of researchers and industry due to their flexibility allowing high-level customization of manufacturing process. However, shared workspace between operators and robots constrains to use specialized industrial robots known as collaborative robots or cobots that have a limited speed and payload characteristics, which are still liable of inducing systematic errors in the positions. The position control of cobots during the design, validation and operation phase is crucial for the human operators’ safety and secure operation to avoid contact with the operator or the surrounding environment, liable of causing damages, as well as for avoiding unnecessary adjustments of the workspace. This work discusses the application of a machine learning-based Digital Twin (DT) including online position correction and error prediction method for HRC systems to allow safe operators’ training and planning of collaborative tasks. Innovatively, a metrologically trustworthy DT is introduced by including calibration of the models and evaluating and propagating the measurement uncertainty of error correction model. The proposed procedure is demonstrated on a state-of-the-art Yaskawa cobot, showing an improvement of positioning precision from 0.970 to 0.216 mm, i.e., of 77.8%.