New Recursive Robust Algorithms for the Estimation of AR Parameters

In order to overcome the difficulty that general robust algorithms are not robust when they are applied to the estimation of autoregressive (AR) parameters in the presence of additive outliers, a new recursive robust estimator for regression and its improved version are proposed in this chapter. The estimator is derived by solving the “normal equation” of the employed cost function using one-step iteration method recursively. The new version is extended from the estimator by modifying the original model conditionally. The simulation results show that the estimator gives a slightly high accuracy than other related estimators while the latter can perform well in nonstationary environments. It is concluded that both of them can suppress the adverse effect of outliers and the latter is more suitable for the estimation of AR parameters.