Minimization of roundoff errors in digital realizations of Kalman filters

A comparison between D. Williamson's method (IEEE Trans. Autom. Contr., vol.AC-30, p.930-9, 1985) and the 0-1 method for the design of the digital Kalman filter is made. The 0-1 method is shown to be much better than Williamson's method and the other methods from the estimation-error pint of view, whereas Williamson's method is slightly better than the 0-1 method from the computation-time point of view. Comparison with all the other transfer-function-based realizations shows that the designer of the digital Kalman filter has to choose only between the 0-1 and Williamson methods.<
>