An ellipsoid algorithm for parameter set estimation

Presented is a recursive ellipsoid algorithm for parameter set estimation of an ARX (autoregressive with exogeneous input) model with bounded disturbances. The algorithm allows for independent, time-varying, upper and lower disturbance bounds, which enables known disturbance bias to be incorporated. Algorithms are presented that allow the application of this ellipsoid algorithm to a special ARMAX (autoregressive moving-average with exogeneous input) model, which allows formal treatment of measurement uncertainty.<
>