SINGLE LAYER PERCEPTRONS NEURAL NETWORKS FOR ONLINE EIGENSTRUCTURE ASSIGNMENT IN MIMO SYSTEMS

Abstract

: To improve the stability performance and the response form of MIMO systems, the control of the modes is relevant to impose the design specifications. A bioinspired method for online eigenstructure assignment for the design of multivariable control systems is presented in this paper. The method is based on artificial neural networks for the execution of the control law and for training the controller gains by state feedback. Aiming at online tuning of state space controllers, the gain matrix that satisfies all the design specifications, a proposal is presented to compute the gain matrices that best meet a given operating range of MIMO dynamic systems. The proposal is evaluated in a mathematical model that represents a fourth order RLC circuit with two input voltages and two controllable voltage levels.