A new approach for instantaneous pole placement with recurrent neural networks and its application in control of nonlinear time-varying systems |
| |
Authors: | Ali Karami-Mollaee Mohammad Reza Karami-Mollaee |
| |
Affiliation: | aElectrical Engineering Department, Tarbiat Modarres University, P. O. Box 14115-143, Tehran, Iran;bElectrical Engineering Department, Mazandaran University, P.O. Box 47135-484, Babol, Iran |
| |
Abstract: | In this paper, we first show that online computation of feedback gain used for pole placement of nonlinear systems in recent years is not reliable, and then we present a new approach for instantaneous pole placement and apply it with dynamical recurrent neural networks for online computation of feedback gain. Because of high-speed convergence of neural network to feedback gain, we can apply this method for pole placement of nonlinear time-varying systems. One strategy for realization of this method is instantaneous linearization, as we do here by simulation. The advantage of the proposed method is a global uniform asymptotical exponential stability (GUAES) of closed-loop system around the equilibrium point. |
| |
Keywords: | Pole placement Recurrent neural network Instantaneous linearization Nonlinear control Global exponential stability |
本文献已被 ScienceDirect 等数据库收录! |
|