An Adaptive Extended Kalman Filter for Non-Gravitational Acceleration Elimination in AHRS

Abstract

Attitude and heading reference system (AHRS) is ubiquitous in the navigation system. Generally, quaternion-based extended Kalman filter (QEKF) is applied to implement attitude estimation by integrating data collected from gyroscopes, accelerometers, and magnetometers. However, accelerometers are sensitive to both gravity and external acceleration, which will reduce the performance of QEKF in dynamic conditions. In order to improve accuracy of attitude estimation, we propose a novel algorithm for non-gravitational acceleration elimination in AHRS. The algorithm detects whether there exists external acceleration which can be used flexibly in a variety of dynamic situations, and then adaptively adjust the noise variance of the extended Kalman filter for compensate external acceleration under dynamic conditions. Experimental results show that the proposed algorithm achieve small average estimation error when compared to standard EKF and the state-of-the-art attitude estimation algorithm in the scenario existing non-gravitational acceleration.