New untuned algorithms for worst case identification

This paper considers the problem of approximating the given set of frequency response samples by an element from a particular user specified subset of H_∞. An untuned algorithm is suggested and a bound is established on the worst case identification error. It is shown that the worst case error is always bounded. Further, if the user specified subset includes the true plant set, the worst case error converges to zero as data becomes infinite and the noise goes to zero. Modified algorithms are proposed which minimise a particular worst case error bound. Finally, a simulation example demonstrates the use of the proposed algorithm