Abstract
In this paper, towards efficient state estimation for large-scale networked linear systems, we propose a method for designing observers to estimate the average behavior of systems, which we call average state observers. First, we give a mathematical formulation of the average state observer. Furthermore, we derive the error system clarifying that not only an initial state estimation error but also the initial value response of systems and the input signal are relevant to the observation error. On the basis of this error analysis, we devise a systematic design procedure for average state observers. The proposed method does not require the information on how to choose a set of average states of systems. The efficiency of the proposed method is shown through a numerical example for a diffusion system.
