Improved least-squares identification for multiple-output non-linear stochastic systems

This study considers the identification problems of multiple-output non-linear equation-error moving average systems.There exist the product items of the parameters between the non-linear and linear parts. To solve this difficulty, the key termseparation technique is adopted. By using the model decomposition technique and the hierarchical identification principle, amaximum likelihood-based recursive extended least-squares estimation algorithm with reduced computational complexity ispresented to estimate the parameters of the non-linear part and the linear part interactively. The simulation results demonstratethe effectiveness of the proposed method.