IMM-UKF based land-vehicle navigation with low-cost GPS/INS

TLDR: Simulations show that compared with the conventional Kalman filtering approach, the IMM-UKF algorithm is more stable and effective, thus improving the convergence speed and accuracy.

Abstract: The motivation of INS/GPS integration is to develop a navigation system that overcomes the shortcomings of each system. Its implementation is essentially based on the filter techniques and error models of INS. If the model changes with the environment, the estimation accuracy is degraded. In this paper, an Interacting Multiple Model Unscented Kalman Filter (IMM-UKF) method was proposed to jointly estimate the position information. This modeling approach makes it possible to employ the UKF to deal with the problem of nonlinear filtering with uncertainty noise. The output of the IMM-UKF is the weighted sum of a bank of parallel unscented Kalman filters. Simulations show that compared with the conventional Kalman filtering approach, the IMM-UKF algorithm is more stable and effective, thus improving the convergence speed and accuracy.