Parameter Identification of Coulomb Oscillator from Noisy Sensor Data

Abstract

Coulomb oscillator is used for analyzing several real-life systems, which demand a precise modeling of the oscillation. The modeling is based on the stochastic estimation of unknown parameters of the model representing the oscillation. This paper introduces a Bayesian approach for the estimation of unknown parameters from sensor-generated noisy data. Among several Bayesian approaches, Gaussian filtering approach is most popular. A major challenge that appeared with the Gaussian filtering is intractable integral, which is approximated numerically. Several Gaussian filters have been reported by using different numerical approximation methods. This paper implements a popular Gaussian filter, named as cubature Kalman filter (CKF), for the estimation of unknown parameters. The CKF uses a third-degree spherical radial rule for the numerical approximation of the intractable integrals. Simulation results conclude a high accuracy of the CKF-based estimate of unknown parameters.