Dynamic Acceleration Compensation for Attitude and Heading Reference System Based on Recurrent Neural Network

Tri-axial magnetometer, accelerometer, as well as gyroscope are commonly used in attitude and heading reference systems (AHRS) to provide three dimensional attitude information for unmanned aerial vehicles (UAV) and other carriers. But the attitude algorithms for AHRS are vulnerable to the external acceleration. To solve this problem, a novel attitude algorithm based on recurrent neural network (RNN) is introduced. Compared to the existing methods, the proposed algorithm shows better robustness against the external acceleration in the experiment on AHRS, and can ensure dynamic attitude accuracy.