This thesis is concerned with the implementation of an adaptive identification algorithm using parallel processing and systolic arrays. In particular, discrete samples o" input and output data of a system with uncertain characteristics are used to determine the parameters of its model. The identification algorithm is based on recursive least squares, QR decomposition, and block processing techniques with covariance resetting. Along similar lines as previous approaches, the identification process is based on the use of Givens rotations. This approach uses the Cordic p technique for improved numerical efficiency in performing the rotations. Additionally, floating point and fixed point arithmetic implementations are compared. Acoessioni For NTIS GRA&I DTIC TAB F] Uuim ou nced j JUS If I a ___ . Avatlabtlity Codes . Avail -and/or . ., Dist Special fyi I

27 Figures and Tables

Cite this paper

@inproceedings{CristiIEIIIIIIIImII, title={IEIIIIIIIImI IIIIIIIII}, author={Roberto Cristi and Angelica Willis} }