A new method in modelling the IMU stochastic errors

摘要

The modelling of the IMU stochastic errors is important in improving the performance of an integrated navigation system. In this paper, we will try to use shaping filter to model the IMU stochastic errors with a unit white noise as the input. The identification of the shaping filters' transfer function from the Allan Variance plot is carefully discussed and analysed. The bias instability is modelled as a summation of two 1st order Markovian processes according to the flicker noise theory. The differential equation and ARMA process based methods can be deduced from the transfer function of the shaping filter using inverse Laplace and Z transform. These two methods show similar coasting performance after implemented in the navigation Kalman filter because they describe the same process from two different aspects. And both of them outperforms the 1st order Markovian process + white noise modelling method. Besides this, the shaping filter also can be used to determine the coefficients of ARMA process and make a bridge between Allan Variance and ARMA process.