Nonlinear Non-Gaussian Filtering Algorithm Based on Cubature Kalman and Particle Filter

Abstract

To resolve the nonlinear non-Gaussian tracking problem effectively, a novel filtering algorithm based on Cubature Kalman Filter (CKF) and Particle Filters (PF) is proposed, which is called Cubature Kalman Particle Filter (CPF). CKF is used to generate the importance density function for PF. It linearizes the nonlinear functions using statistical linear regression method through a set of Gaussian cubature points. It need not compute the Jacobian matrix. Moreover, it makes efficient use of the latest observation information into system state transition density, thus greatly improving the filter performance. The simulation results show that CPF has higher estimation accuracy and less computational load comparing against the widely used Unscented Particle Filter (UPF).