Fast and stable recursive algorithms for continuous-time and discrete- time model conversions (Q1103336)

Consider a continuous-time system \(\dot x=Ax+Bu\) and its equivalent discrete-time model \(x(k+1)=Gx(k)+Hu(k)\) for a fixed sampling period \(T<\pi /\omega_ s\), where \(\omega_ s\) is the highest frequency component of continuous-time model. The non-recursive algorithms such as the scaling, squaring, continued-fraction method for finding A, B from G, H and the square root, scaling, continued-fraction method for finding G, H from A, B are summarized including their truncation errors. New rapidly convergent and numerically stable algorithms for finding A from G and G from A are derived. These algorithms are based on the relaxation of the constraints imposed on the existing model conversion algorithms. The numerical stability of both algorithms is proved. An example is supplied.