A comparison of discretization methods for parameter estimation of nonlinear mechanical systems using extended kalman filter: Symplectic versus classical approaches

This paper presents two symplectic discretization methods in the context of online parameter estimation for nonlinear mechanical systems. The symplectic approaches are compared to established discretization methods (e.g. Euler Forward and Runge Kutta) regarding accuracy and computational effort. The methods are compared using two mechanical simulation models of a real belt-drive system: a nonlinear two-mass system with two degrees of freedom and a nonlinear three-mass system with three degrees of freedom. In addition, the influence of the discretization method on the performance of an augmented Extended Kalman Filter (EKF) estimating the parameter of the two-mass system is analyzed. The simulation shows improved accuracy of the calculated discrete-time solution using symplectic integrators in comparison to the conventional methods, with almost the same or lower computational cost. Additionally, the parameter estimation based on the EKF in combination with the symplectic integration scheme leads to more accurate values.