Affiliations: [a] Departamento de Mecatrónica, Escuela de Ingeniería y Ciencias, Tecnologico de Monterrey, Carr. al Lago de Guadalupe Km.3.5, 52926, Estado de México, México | [b] Departamento de Ingeniería Eléctrica, Cinvestav-IPN, Av. IPN 2508, A.P. 14-740, 07300, Ciudad de México, México
Correspondence:
[*]
Corresponding author. Angel Ramírez-Martínez, Departamento de Mecatrónica, Escuela de Ingeniería y Ciencias, Tecnologico de Monterrey, Carr. al Lago de Guadalupe Km.3.5, 52926, Estado de México, México. Tel.: +52 55 5864 5536; E-mail: [email protected].
Abstract: This paper presents a data-driven control approach for a five-bar robot with compliant joints. The robot consists of a parallel mechanism with compliant elements that introduce uncertainties in modeling and control. To address this fact, it is implemented a model-less data-driven controller based on a Feedforward Neural Network Module (FNNM) that identifies the inverse dynamics of the robot. The FNNM is incorporated into a coordination of Feedforward Control Method (CFCM) to achieve precise trajectory tracking. Experiments compare the compliant joints robot to a bearing-joint robot performing pick-and-place tasks from 0.15 to 3.15 Hz. Results show the compliant robot maintaining trajectory tracking up to 1.25 Hz with a Root Mean Square Error (RMSE) of 9.02 mm.
