Low-Cost MEMS-Based AHRS Using an Adaptive Robust Error-State Kalman Filter for Motion Capture

MEMS-based AHRS systems are increasingly applied to human motion capture tasks. Their portability and low cost allow for deployment in various indoor and outdoor environments. However, the estimated attitude obtained from MEMS-based AHRS systems is susceptible to environmental degradation from magnetic disturbances. This paper proposes an Adaptive Robust Error-State Kalman Filter (ARESKF) to enhance attitude estimation in complex magnetic disturbance environment. Unit quaternions and rotation vectors are employed to propagate the nominal state and the error state, respectively. During the measurement update, the accuracy of the estimated attitude is improved by the adaptive adjustment of the associated measurement noise covariance matrix. A novel magnetic disturbance detector is designed to classify magnetic field environments, while an innovative robust switcher dynamically adjusts the magnetic reference vector based on varying magnetic field conditions. These techniques enable adaptive switching of the magnetic reference vector during the Kalman filter's measurement update based on the type of magnetic field vector detected. This approach improves stability and accuracy of heading estimation in comprehensive magnetic disturbance environments. Through simulations and experiments, the proposed ARESKF algorithm has been compared with five other attitude estimation algorithms, demonstrating its superior performance in attitude estimation.