TY - GEN
T1 - Framework for discrete-time model reference adaptive control of weakly nonlinear systems with HONUs
AU - Benes, Peter M.
AU - Bukovsky, Ivo
AU - Vesely, Martin
AU - Voracek, Jan
AU - Ichiji, Kei
AU - Homma, Noriyasu
N1 - Funding Information:
Acknowledgements Authors acknowledge support from the EU Operational Programme Research, Development and Education, and from the Center of Advanced Aerospace Technology (CZ.02.1.01/0.0/0.0/16_019/0000826), and the Japanese JSPS KAKENHI Grant Number 15J05402.
Publisher Copyright:
© Springer Nature Switzerland AG 2019.
PY - 2019
Y1 - 2019
N2 - This paper reviews the Higher Order Nonlinear Units (HONUs) and their fundamental supervised sample-by-sample and batch learning algorithms for data-driven controller learning when only measured data are known about the plant. We recall recently introduced conjugate gradient batch learning for weakly nonlinear plant identification with HONUs and we compare its performance to classical Levenberg-Marquard (LM). Further, we recall recursive least square (RLS) adaptation and compare its performance to L-M learning both for plant approximation and controller tuning. Further, a model reference adaptive control (MRAC) strategy with efficient controller learning for linear and weakly nonlinear plants is proposed with static HONUs that avoids recurrent computations, and its potentials and limitations with respect to plant nonlinearity are discussed. Recently developed stability approach for recurrent HONUs and for closed control loops with linear plant and nonlinear (HONU) controller is recalled and discussed in connotation stability of the adaptive closed control loop.
AB - This paper reviews the Higher Order Nonlinear Units (HONUs) and their fundamental supervised sample-by-sample and batch learning algorithms for data-driven controller learning when only measured data are known about the plant. We recall recently introduced conjugate gradient batch learning for weakly nonlinear plant identification with HONUs and we compare its performance to classical Levenberg-Marquard (LM). Further, we recall recursive least square (RLS) adaptation and compare its performance to L-M learning both for plant approximation and controller tuning. Further, a model reference adaptive control (MRAC) strategy with efficient controller learning for linear and weakly nonlinear plants is proposed with static HONUs that avoids recurrent computations, and its potentials and limitations with respect to plant nonlinearity are discussed. Recently developed stability approach for recurrent HONUs and for closed control loops with linear plant and nonlinear (HONU) controller is recalled and discussed in connotation stability of the adaptive closed control loop.
KW - Conjugate gradients
KW - Higher order neural units
KW - Model reference adaptive control
KW - Nonlinear dynamics
KW - Polynomial neural networks
UR - http://www.scopus.com/inward/record.url?scp=85067250372&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85067250372&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-16469-0_13
DO - 10.1007/978-3-030-16469-0_13
M3 - Conference contribution
AN - SCOPUS:85067250372
SN - 9783030164683
T3 - Studies in Computational Intelligence
SP - 239
EP - 262
BT - Computational Intelligence - 9th International Joint Conference, IJCCI 2017, Revised Selected Papers
A2 - Madani, Kurosh
A2 - Merelo, Juan Julian
A2 - Warwick, Kevin
A2 - Sabourin, Christophe
A2 - Warwick, Kevin
PB - Springer Verlag
T2 - 9th International Joint Conference on Computational Intelligence, IJCCI 2017
Y2 - 1 November 2017 through 3 November 2017
ER -