An efficient VLSI implementation of real-time Kalman filter

An efficient systolic implementation for a real-time Kalman filter is presented. The filter is implemented by propagating the error information matrix. The whole computation for each estimate is done in five processing steps instead of eight by exploiting inherent parallelism in the algorithm. This is achieved by dividing the eight steps into two groups, each of which can be computed simultaneously on a systolic array, thereby reducing the total computation time by 40%. Each processing step follows a modified Faddeev´s algorithm which is implemented on a bitrapezoidal array. Using a bitrapezoidal array doubles the throughput obtained by a single trapezoidal array. Techniques used for data skewing and storage organization are efficient, thus reducing the silicon area and complexity of control, and increasing the speed of computation