Sliding Window Iterative Identification for Nonlinear Closed-Loop Systems Based on the Maximum Likelihood Principle

The parameter estimation problem for the nonlinear closed-loop systems with moving average noise is considered in this article. For purpose of overcoming the difficulty that the dynamic linear module and the static nonlinear module in nonlinear closed-loop systems lead to identification complexity issues, the unknown parameters from both linear and nonlinear modules are included in a parameter vector by use of the key term separation technique. Furthermore, an sliding window maximum likelihood least squares iterative algorithm and an sliding window maximum likelihood gradient iterative algorithm are derived to estimate the unknown parameters. The numerical simulation indicates the efficiency of the proposed algorithms.