Robust Optimization Algorithm for Attitude Estimation Based on Multisensor Fusion Under Magnetic Disturbance Conditions

Wearable devices face significant challenges in indoor complex magnetic field environments, especially the problem of disturbance in the accuracy and stability of attitude estimation by inertial measurement units (IMUs). This article proposes a method for detecting magnetic disturbance and calculating the trend of magnetic field change by processing magnetic field data. The method is designed to identify and classify different change trends, thereby informing the decision of whether to fuse magnetometer data during attitude estimation. Furthermore, this article fuses the advantages of the gradient descent algorithm (GDA) and the Gauss–Newton algorithm to propose a hybrid optimization algorithm for attitude estimation, thereby enhancing the algorithm’s accuracy. Additionally, it employs a dynamic adjustment method to adaptively adjust the weights of the two algorithms in different environments, thereby improving the algorithm’s robustness. The experimental results show that compared with the mainstream extended Kalman algorithm, the proposed method in this article improves the root mean square error (RMSE) of Roll, Pitch, and Yaw by 58.01%, 66.15%, and 90.51%, respectively. Compared to other disturbance-resistant algorithms, it improves 66.69%, 65.10%, and 49.23% on the RMSE of Roll, Pitch, and Yaw, respectively. In addition, the improvement in accuracy and stability of the proposed method in this article is further verified by boxplot analysis.