TY - JOUR
T1 - A Novel Inverse Kinematic Solution of a Six-DOF Robot Using Neural Networks Based on the Taguchi Optimization Technique
AU - Ibarra-Pérez, Teodoro
AU - Ortiz-Rodríguez, José Manuel
AU - Olivera-Domingo, Fernando
AU - Guerrero-Osuna, Héctor A.
AU - Gamboa-Rosales, Hamurabi
AU - Martínez-Blanco, Ma del Rosario
N1 - Publisher Copyright:
© 2022 by the authors.
PY - 2022/10
Y1 - 2022/10
N2 - The choice of structural parameters in the design of artificial neural networks is generally based on trial-and-error procedures. They are regularly estimated based on the previous experience of the researcher, investing large amounts of time and processing resources during network training, which are usually limited and do not guarantee the optimal selection of parameters. This paper presents a procedure for the optimization of the training dataset and the optimization of the structural parameters of a neural network through the application of a robust neural network design methodology based on the design philosophy proposed by Genichi Taguchi, applied to the solution of inverse kinematics in an open source, six-degrees-of-freedom robotic manipulator. The results obtained during the optimization process of the structural parameters of the network show an improvement in the accuracy of the results, reaching a high prediction percentage and maintaining a margin of error of less than 5%.
AB - The choice of structural parameters in the design of artificial neural networks is generally based on trial-and-error procedures. They are regularly estimated based on the previous experience of the researcher, investing large amounts of time and processing resources during network training, which are usually limited and do not guarantee the optimal selection of parameters. This paper presents a procedure for the optimization of the training dataset and the optimization of the structural parameters of a neural network through the application of a robust neural network design methodology based on the design philosophy proposed by Genichi Taguchi, applied to the solution of inverse kinematics in an open source, six-degrees-of-freedom robotic manipulator. The results obtained during the optimization process of the structural parameters of the network show an improvement in the accuracy of the results, reaching a high prediction percentage and maintaining a margin of error of less than 5%.
KW - backpropagation
KW - inverse kinematics
KW - optimization methods
KW - robotics
UR - http://www.scopus.com/inward/record.url?scp=85139972136&partnerID=8YFLogxK
U2 - 10.3390/app12199512
DO - 10.3390/app12199512
M3 - Artículo
AN - SCOPUS:85139972136
SN - 2076-3417
VL - 12
JO - Applied Sciences (Switzerland)
JF - Applied Sciences (Switzerland)
IS - 19
M1 - 9512
ER -