The extended Kalman filter (EKF) is widely used for the integration of measurements from the Global Positioning System (GPS) and inertial navigation system technologies. The EKF is applied to nonlinear systems via a linearization of the system, and is reliable for weakly linear systems.
For strong linear systems however, the EKF sometimes suffers from divergence. To overcome this shortcoming the so-called sigma-point Kalman filter (SPKF) has been proposed. The SPKF uses a finite number of sigma points to implement the nonlinear transformations directly instead of linearizing them.
This paper applies the SPKF to the initial alignment problem, because it could compensate for the influence of linearization error on Microelectromechanical system Intertaial measurement units (MEMS-IMU) calculations.
The nonlinear error models for a MEMS Inertial Measurement Unit (IMU) are derived by representing the orientation in the quaternion. Both simulation and field test results are presented to demonstrate the state estimation performance of the SPKF. The simulation results show that the SPKF converges faster than the EKF for the same filter precision.
The state estimation precision of the two filters is compared – the covariance estimation accuracy of SPKF is better than that of EKF. These advantages make the SPKF suitable for fast and highly maneuverable trajectories.
To read this external content in full, download the complete paper from the authors’ article archive at the University of New South Wales.