Nonlinear differential equations: Parametric identification by exact polynomial spline schemes (Q1280793)

The identification problem of a continuous-time nonlinear state-space vector differential equation from output observations of an adequate dynamical system is considered. It is assumed that system state variables can be exactly represented in the polynomial spline form and a recursive computing procedure for establishing spline parameters is presented. Necessary and sufficient conditions of efficiency of the proposed routine are given and an approximation error at each step of the procedure is evaluated. It is shown that the proposed algorithm attains the best possible approximation accuracy in the spline space. Using the spline state-vector solutions, the problem of parametric identification of a complex exponential-type dynamic system, being a collection of homogeneous components, is next solved by employing trajectory averaging and a state-equation linearization technique. An example presenting the efficiency of the method is included.