TY - GEN
T1 - RBF Neural Network Based on FT-Windows for Auto-Tunning PID Controller
AU - Castro, O. F.Garcia
AU - Velasco, L. E.Ramos
AU - Navarrete, M. A.Vega
AU - Rodriguez, R. Garcia
AU - Mayorga, C. R.Domínguez
AU - Hernández, E. Escamilla
AU - Moreno, L. N.Oliva
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - The weighted function windows are used in many areas as signal analysis and application systems. In addition, the weighted functions are broad uses in filter design where different windows allow to choose different filter characteristics. The most common individual window types are rectangular, Hanning, Flat Top, and Keiser-Bessel. This paper presents the Flat-Top Windows (FTW) applied to control systems where the FTW are used as activation functions on a radial basis neural network (RBF). Contrary to the "traditional" FT weighted function windows, where time windows limit the information, this paper proposes new ones that, including new parameters, allow translation and dilation of the window. Additionally, these new parameters are updated using a gradient descent algorithm. The new FTW is applied to the Quanser helicopter control where the RBF neural network is used for: a) the input-output identification of the system and b) auto-tuning PID controllers. Numerical simulation results are presented to show the system’s performance under different conditions.
AB - The weighted function windows are used in many areas as signal analysis and application systems. In addition, the weighted functions are broad uses in filter design where different windows allow to choose different filter characteristics. The most common individual window types are rectangular, Hanning, Flat Top, and Keiser-Bessel. This paper presents the Flat-Top Windows (FTW) applied to control systems where the FTW are used as activation functions on a radial basis neural network (RBF). Contrary to the "traditional" FT weighted function windows, where time windows limit the information, this paper proposes new ones that, including new parameters, allow translation and dilation of the window. Additionally, these new parameters are updated using a gradient descent algorithm. The new FTW is applied to the Quanser helicopter control where the RBF neural network is used for: a) the input-output identification of the system and b) auto-tuning PID controllers. Numerical simulation results are presented to show the system’s performance under different conditions.
KW - Autotuning PID controller
KW - FT windows
KW - Helicopter model
KW - RBF neural network
UR - http://www.scopus.com/inward/record.url?scp=85142638338&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-19493-1_11
DO - 10.1007/978-3-031-19493-1_11
M3 - Contribución a la conferencia
AN - SCOPUS:85142638338
SN - 9783031194924
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 138
EP - 149
BT - Advances in Computational Intelligence - 21st Mexican International Conference on Artificial Intelligence, MICAI 2022, Proceedings
A2 - Pichardo Lagunas, Obdulia
A2 - Martínez Seis, Bella
A2 - Martínez-Miranda, Juan
PB - Springer Science and Business Media Deutschland GmbH
T2 - 21st Mexican International Conference on Artificial Intelligence, MICAI 2022
Y2 - 24 October 2022 through 29 October 2022
ER -