IMU Attitude Estimation Method Based on Robust Adaptive Kalman Filter

Aiming at addressing the shortcomings of high noise and low accuracy in inertial measurement units (IMUs) and the low accuracy of conventional attitude estimation algorithms, a fusion algorithm combining a dynamic detection model with robust adaptive Kalman filtering (RAF) is proposed. The algorithm is based on the dynamic detection model utilizing Principal Component Analysis (PCA) and Mahalanobis Distance (MD), which reduces the influence of IMU multidimensional noise, improves the correlation between the data, and effectively captures the key features of attitude change. The data, after feature approximation using the PCA-MD algorithm, is then imported into the RAF model to obtain the attitude information of the IMU. Experimental results demonstrate that the fusion algorithm combining PCA, MD, and RAF can suppress attitude noise and drift in static conditions, as well as mitigate the impact of perturbations in dynamic conditions, thereby effectively improving the accuracy of attitude estimation.