A Stochastic Aspect of Decentralized Control

Abstract

In this paper we aim to calculate a decentralized dynamic output feedback regulator for the linear multivariable stochastic process. Effectively the classical seperation principle does not hold to this structure of control, so we use a direct search method to verify our aim.With the decentralized control structure we try to get a satisfactory suboptimal performance which may represent a manner to avoid some of the criticisms mentioned by Variya (ref.2) about the hierarchical control. Also we avoid the construction of decentralized estimators in which the control agents make complete estimations of the 'state vector.The dynamic output feedback regulators fed by partial information of the output vector are constructed by a direct search algorithm which defines the values of the regulator parameters in the finite time also in the infinite time cases. The noise-free output and the static regulators are special cases of the considered structure. Finally a numerical example is treated to present the utility of the algorithm.