Robust recursive estimation for correlated observations

The Kalman filter is probably the most popular recursive estimation method. It is, however, known to be non-robust to spuriously generated observations. Much attention has been focused on finding the so-called robust recursive estimation under the assumption that the observations are independent. In this paper, we show that Lin and Guttman's robust recursive estimation scheme can be easily applied to the correlated observations. Examples when the noise follows an AR(2) process with/without outliers are given for illustration.