Unscented particle filter based Gaussian process regression for IMU/BDS train integrated positioning

Aiming at the inaccurate of system dynamic model may reduce the filtering effect, a new Unscented Particle Filter based on improved Gaussian process (GPUPF) is proposed in this paper. The importance density function of UPF is obtained by Gaussian process regression, when the system model and noise are uncertain, GPR is used to revise and estimate the system, determine the covariance matrices of system noises, gets better importance density function, and enhance the adaptive capability of UPF. The improved algorithm was applied to the IMU/BDS train integrated positioning system. Simulation results show that the proposed algorithm is better than standard UPF, leading to improved positioning precision.