Estimating model parameters from noisy observations for nonlinear dynamical systems

An optimization-based parameter estimation approach for nonlinear dynamical systems subject to noisy observations is proposed. Spline and numerical differentiation methods are used to smooth noisy observations and to estimate the time derivative of the dynamical system, respectively. Subsequently, the parameter estimation problem can be reduced to a standard least-square problem or a linear programming problem. Two numerical examples, the Lorenz system and a time-delay chaotic system, are illustrated for verifying the performance of the proposed approach. Simulations have shown that the proposed approach is fast and robust to noise. © 2012 Copyright Taylor and Francis Group, LLC.