Partially coupled estimation algorithm for discrete-time multiple-input multiple-output systems

In this paper, a partially coupled recursive least squares (PC-RLS) algorithm is proposed to estimate the parameters of a class of multivariable systems. The first step is to decompose the identification model into submodels or subsystems, resulting in the subsystem least squares (S-RLS) algorithm. For the same parameters contained in each submodel, the PC-RLS algorithm is to couple the estimation of the common parameters of each submodel one by one, while the estimates of the common parameters are mutually independent for the S-RLS algorithm. The analysis shows that the PC-RLS algorithm requires less computational burden than the standard recursive least squares method and the S-RLS method. The simulation tests verify the effectiveness of the proposed algorithm.