Interdependent multiobjective control using Biased Neural Network (Biased NN)

Abstract

A Biased Neural Network (Biased-NN) is proposed to solve an interdependent multiobjective control problem. The main idea of the Biased-NN stems from a decoupled fuzzy sliding mode control scheme that provides a simple way to achieve asymptotic stability for a class of decoupled systems. Each neuron in the Biased-NN is used to approximate a sign function in order to replace the sliding mode control structure with the Biased-NN. Such a feature is useful for handling the interdependent multiobjective control problem based upon the proposed supporting strategy. While previous works require a priori knowledge for all the objectives, the proposed method uses only expert knowledge of the objective that is considered the main concern. Simulations are conducted to show the effectiveness of the Biased-NN.