An important part of optimal estimation technology, the Kalman filter is a computationally intensive application that has been limited either to non-real time realizations or to realizations that can afford vast amounts of mainframe hardware. The potential use of the Kalman filter theory could be greatly enhanced by a low cost, high performance machine capable of computing the recursive matrix equations in real time.
The use of pipelined parallel architectures allows the Kalman filter equations to be realized with much greater efficiency than previous implementations. A reconfigurable, few instruction, multiple data, orthogonal, pipelined, systolic array processor will be used to implement the recursive algorithm of the filter. Since the architecture is reconfigurable, a single systolic array will perform all of the required operations. The architecture selected provides a general foundation for other applications involving matrix computations to build upon.
A previously designed algorithm for pipelined matrix multiplication is employed, and a modified version of an inversion algorithm which is based on Cholesky's method is used. The resulting system improves the performance of the Kalman filter by about a factor of three over an implementation by Liu and Young.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu
Papadourakis, George M.
Master of Science (M.S.)
College of Engineering
Length of Campus-only Access
Masters Thesis (Open Access)
Bapst, Mark V., "A Reconfigurable Orthogonal Systolic Array Implementation of a Kalman Filter" (1987). Retrospective Theses and Dissertations. 5087.
Contributor (Linked data)