Quadrature Kalman Particle Filter
Chongzhao Han · Xi'an Jiaotong Daxue xuebao · 2009
A new kind of quadrature Kalman particle filter is proposed for the state estimation of nonlinear/non-Gaussian systems.The new algorithm uses the quadrature Kalman filter(QKF) to generate the importance density function,and linearizes the nonlinear functions using the statistical linear regression method through a set of Gaussian-Hermite quadrature points.The algorithm does not evaluate the Jacobian matrix,and is easy to implement.Moreover,the importance density function integrates the latest observations into the system state transition density,so that the approximation to the system posterior density is improved.Theoretical analysis and experimental results show that,compared with the unscented particle filter(PF-UF),the estimation accuracy of the new particle filter is improved by almost 18%,and its calculation cost is slightly reduced,which indicates the PF-QKF to be an effective nonlinear filtering algorithm.