UBC Theses and Dissertations
Linear systems identification algorithms for a minicomputer Zuercher, Heinrich
Two methods for identifying multiple-input mutliple-output linear time-invariant discrete systems from input-output data are derived. The selector matrix principle of Gopinath is used and two special classes of selector matrices that lead to canonical system models are introduced. Very general input sequences can be applied. However, a few restrictions exist for the initial system state. The input matrix is identified columnwise. Both methods are considerably better than Gopinath's method in terms of storage requirements and numerical accuracy. Tests on a large computer are performed. One method is implemented on a minicomputer with good results.
Item Citations and Data