A recursive parameter identification algorithm for nonlinear errors-in-variables models

Recursive identification of model parameters from data is an essential tool in adaptive signal processing and adaptive control applications. For general nonlinear regression problems where errors are present in both dependent and independent variables, the errors-in-variables (EIV) model is frequently employed to avoid a biased estimation of the system parameters. This work presents a recursive identification algorithm for nonlinear EIV models which is derived from an offline sequential quadratic programming solution. A proof of convergence is provided and the performance of the proposed algorithm is illustrated on benchmark simulation examples. The results demonstrate the numerical efficiency of the recursive algorithm, as well as its ability to track time-varying system parameters.