Filtering based least squares parameter estimation algorithms for Hammerstein nonlinear CARMA systems

Hammerstein system identification is difficult because there exist the product items of the parameters between the static nonlinear block and the dynamic linear block. This paper presents a novel parameter separation based recursive least squares (PS-RLS) identification algorithm for resolving this problem. Its basic idea is to use a linear filter to filter the output data and the noise, and then to obtain two new identification submodels in each of which the output is linear in the corresponding parameter vector. Compared with the over-parametrization based recursive least squares method, the proposed algorithm can avoid estimating the redundant parameters and has a higher computational efficiency. The simulation results show that the proposed PS-RLS algorithm can generate highly accurate parameter estimates with less computational effort.