Combined Quaternion-Based Error State Kalman Filtering and Smooth Variable Structure Filtering for Robust Attitude Estimation

Abstract

This paper presents a novel robust quaternion-based error state Kalman filter (ESKF) for coping with modeling uncertainty in inertial measurement unit (IMU)-based attitude estimation. The smooth variable structure filter (SVSF) has recently been proposed and proven to be robust to modeling uncertainty. In an effort to combine the accuracy of an ESKF with the robustness of the SVSF, the ESKF and SVSF algorithms have been merged to create the ESKF-SVSF algorithm. In particular, a comprehensive fault detection strategy has been proposed to combine the optimality of the ESKF and the robustness of the SVSF. The proposed ESKF-SVSF algorithm was validated on experimental data collected from a small unmanned aerial vehicle (UAV) in the presence of faulty gyroscope signals. In the experiment, four faulty test cases were consideblack, involving the injection of two types of faults into the raw gyroscope signals to simulate modeling uncertainty. Although the proposed ESKF-SVSF algorithm incurs a slightly increased computational load, the experimental results demonstrate that the proposed algorithm yields more accurate attitude estimates than the conventional approach does in the presence of modeling uncertainty.