A Convolutional Neural Network-Based Parametric Identification for the Dynamic Model of a Cartesian Robot in a FPGA Framework

Parametric identification in robotic systems involves determining key parameters within a system's model to enhance performance. This paper presents a novel convolutional neural network (CNN) architecture designed for a parametric identification of four parameters in the dynamic model of a Cartesian robot with three degrees of freedom, which it was designed by the authors of this paper. The dynamic model is derived by the lumped parameter methodology using the Euler-Lagrange equations of motion. Detailed explanations of the neural network's configuration and training methodology are provided, aiming to maximize its effectiveness in real-world applications, thereby improving overall performance and robustness. The results show that the parametric identification of the CNC robot achieves a similarity percentage of approximately 85 % with the experimental data.