Fast Numerical Method Applied to System Modeling and State Estimation

Abstract The philosophy of a numerical method, especially suited to fast repetitive computations is presented, together with the main results of corresponding theoretical derivation, previously reported by the author. This approach uses a sharp easy-to-implement formalism in order to provide an off-off-line numerically stable computation, such that an optimal trade-off between fast memory requirements, speed and accuracy of the subsequent off-line commitation is made, by saving appropriately chosen matrices. A microprocessor implementation of computed discrete time approximation of a continuous-time minimal order observer, and the computer-aided design of both observer and control law as well use the high speed algorithm. The identification of a simplified model with discrete varying structure of a nonlinear plant is a second example outlined in this paper.