Worst Case Identification

Paresh Date (CUED)

There are two objectives in the worst case/deterministic identification problem:

To map the experimental frequency response data into a stable plant transfer function;

To provide an upper bound on the worst case identification error, in terms of a priori information about the plant and the measurement noise.

In this talk, the conventional worst case identification algorithms will be outlined. Then some new algorithms will be introduced, which map the given set of frequency response samples to a particular, user specified subset of H-infinity. The convergence properties of the proposed algorithms are compared with those of the conventional ones. The use of these algorithms is illustrated by simulation examples. One family of new algorithms is shown to be optimal , in a certain sense, under restricted model order.

Back to Control Seminars Page