Orientation Estimation by Partial-State Updating Kalman Filter and Vectorial Magnetic Interference Detection
Maoran Zhu, Wei Ouyang, Yuanxin Wu
Abstract
Attitude estimation based on a magnetic and inertial measurement unit is popular in the areas of robotics, aerospace, and unmanned aerial vehicles. Nevertheless, the magnetometer is prone to external magnetic disturbance that deteriorates the state estimation. This article proposes a novel partial state update strategy in the extended Kalman filter (PS-EKF) by optimally solving the posteriori covariance optimization problem under the Kalman gain constraint, so as to obviate the magnetic measurement update on two level angles and the inertial sensor biases. Besides, a new magnetic disturbance detection technique (vectorial detector) considering both the norm and the direction of the magnetometer measurement is proposed. Simulations and experiments show that the proposed PS-EKF method can be totally immune to the magnetic interference on the level angles and sensors biases, in contrast to the traditional extended Kalman filter using either the direct magnetic output or the indirect yaw as the measurement.