Improvement of Attitude Estimation Using Hidden Markov Model Classification

Abstract

This paper introduces the attitude estimation method of AHRS using an Extended Kalman Filter(EKF) with a filter tuning algorithm based on Hidden Markov Model(HMM). The AHRS uses inertial sensors and magnetometers to calculate its attitude. It is known that the attitude update using gyros are prone to diverge and hence the attitude error needs to compensate using accelerometers and magnetometers. In this paper, a Kalman filter model with a state variables represented by a quaternion is presented and a model changing algorithm is used to make the filter more robust to acceleration and magnetic disturbances. If the AHRS measures any disturbances which are caused by movement of the vehicle, HMM estimates the existence of disturbances. Using these estimates HMM changes the filter gain using tuning parameters of the filter. Results of EKF tuned by HMM indicate that the proposed method makes robust to disturbances more properly.