A deep learning approach to identify dynamic parameters in a cartesian robot

This article proposes a deep-learning approach to identify dynamic parameters in a Cartesian robot. The parametric identification of a robot allows us to know its dynamic model. The dynamic model will enable us to perform control the robot in a better way. Two structures of convolutional neural networks generated with a platform that assists in developing previously developed neural networks are proposed. The reconstruction of the pair is compared with the parameters identified by both networks to find the best one for this purpose. The construction of the pair using simulated data reaches a similarity of 99.39% and using experimental data 95.32%.