DCM-based orientation estimation using cascade of two adaptive extended Kalman filters

Abstract

In this paper, we propose an effective algorithm to estimate orientation angles (roll, pitch, and yaw) from Inertial Measurement Unit (IMU). This algorithm uses two adaptive extended Kalman filters (AEKF) to estimate the Direction Cosine Matrix (DCM), the external acceleration and the magnetic disturbance. First 6-state filter estimates three elements in the third row of the DCM and the external acceleration on three axes. The second one estimates three elements in the first row of the DCM and the magnetic disturbance on three axes. The last three elements of the DCM are computed from the DCM orthogonalization property. This method overcomes original problems when IMU is moved by external acceleration, and it is disturbed in magnetic environment. In addition, it helps reduce effort on computation from 15-states filter to two 6-states filters. The performance of proposed algorithm is verified by applying it to a 9-DOF IMU and testing its accuracy in various conditions. Experiment results have shown that the proposed algorithm achieves accurate estimation of orientation. The RMS error of all three angles is less than 2°.