Two new estimation algorithms for linear models with unknown but bounded measurement noise

Attention is given to linear systems described by y=A/sub theta /+e where the measurement error vector e is unknown but bounded. Two algorithms for sequential parameter identification are introduced. Their convergence properties are illustrated and compared with those of existing algorithms. A simulation study is carried out using simulated data to investigate the possible practical use of the algorithms. Their performances are compared with those of other offline algorithms as well as with those of the widely used least-squares estimates.<
>