Multi-Steps Weighted ARMA Identification Algorithm for the Multi-sensors System with Unknown Parameters

In this paper, we describe the formatting guidelines for For the multi-sensors time-invariant system with unknown parameters, in order to improve the accuracy of the identification, based on the ARMA model, a kind of multi-steps identification algorithm is presented. Step a: we use recursive extended least squares method to get the local estimates of the unknown parameters, we use arithmetic mean to get the first fusion estimates of the parameters. Step b: we use correlated function method to get the estimates of observation noises of every single sensor. Step c is the key step of our algorithm which differs from the conventional 2-steps algorithm. Step c: we take the related information of the estimates of every single sensor as the weight, we take the weighted fusion of the local estimates as the final estimates of the parameters. Compared to the real values, the final estimates are more accurate than the first estimates. We use Matlab to simulate a typical example, the simulation results show that the final estimates have better convergence than the first estimates, which could show the accuracy advantage of the multi-steps identification algorithm presented in this paper.