Output error identification is not globally stable in the 'nonideal' case
暂无分享,去创建一个
The authors demonstrate that the parameter estimates of the output error identification algorithm are not necessarily globally bounded when the prediction error is corrupted by an arbitrarily small disturbance. The cause of the parameter estimate drift to infinity is an input which can be interpreted as an open-loop control to an unstable system. The disturbance could actually be vanishing.<<ETX>>
[1] Robert R. Bitmead,et al. Persistence of excitation conditions and the convergence of adaptive schemes , 1984, IEEE Trans. Inf. Theory.
[2] C. Richard Johnson,et al. Adaptive IIR filtering: Current results and open issues , 1984, IEEE Trans. Inf. Theory.