An enhanced systolic architecture for recursive least squares applications.

摘要

This thesis presents an enhanced systolic architecture based on the Matrix Update Systolic Experiment (MUSE) recently developed at Massachusetts Institute of Technology, Lincoln Laboratory. The MUSE is a linear systolic array for adaptive nulling of a 64 degree of freedom sidelobe canceller radar. The architecture performs Cholesky factorization of the system covariance matrix using Givens rotations. A fixed sidelobe canceller constraint is used to solve for the adaptive weights.;An enhanced MUSE architecture is presented capable of solving general least squares problems with arbitrary constraints. A method for back solving the Cholesky factor using only the systolic array processing elements is presented. In addition, it has been shown how the architecture can be modified to perform hyperbolic downdating of the Cholesky matrix. The performance of the enhanced architecture is examined both by analytical means and computer simulations.