Adaptive Neural Control with Integral-Plus-State Action
Ieroham Baruch*, Alfredo Martinez*, Boyka Nenkova**
*CINVESTAV-IPN, Ave. IPN No 2508, A.P. 14-470 Mexico D.F., C.P. 07360
** Institute of Information Technologies, 1113 Sofia
Abstract: An indirect adaptive neural control with Integral-Plus-State (IPS) action, is proposed. The control scheme contain one identification and state estimation Recurrent Trainable Neural Network. The identified plant parameters and the estimated state vector are used to compute an adaptive IPS control. Two control schemes are proposed, containing one or two integrals in the control law. The good tracking abilities of this adaptive IPS control are confirmed by simulation results, obtained with a mechanical plant with friction model. Copyright © 2002 IFAC.
Keywords: Neural Networks, Backpropagation, PID Controllers, Adaptive systems, Reference signal, Feedback Control, Mechanical systems, Friction.