Worst case identification using FIR models

This paper considers a robustly convergent algorithm for worst case identification using FIR models. A new and stronger notion of robust convergence is established, and error bounds are obtained for a fixed model order as the length of data tends to infinity. The algorithm is shown to be implementable as a solution to an LMI optimisation problem. A robustly convergent algorithm for identification of plant coprime factors is suggested. An iterative technique for identification in the ν-gap metric is given. Two simulation examples demonstrate the use of these algorithms.