Efficient ARX identification algorithms with full parallelism

A highly parallel algorithm for ARX system identification is developed. It involves only scalar operations which are of the Schur-type and can be implemented by inner product processors of the multiply and add form. A by-product of the above development is a Levinson-type algorithm in transversal form. Modularity of the arrays, local communication, and pipelineability make them suitable for VLSI implementation.<
>