\(H_\infty\)-robustness of adaptive filters against measurement noise and parameter drift (Q1307118)

The paper studies the \(H_\infty\) performance of three classical adaptive filtering algorithms (normalized least mean squares, Kalman filter and central \(H_\infty\) filter) with respect to both measurement noise and parameter drift. It is shown that a compromise between \(H_\infty\) and \(H_2\) performance criteria is necessary to guarantee a satisfactory behaviour of the filtering algorithms. Numerical examples show that the central \(H_\infty\) filter is possibly the best suited to achieve a satisfactory tradeoff between the two criteria.