Attitude and Heading Adaptive Estimation Using a Data Driven Approach

Accurate attitude and heading is required for many applications. For pedestrian navigation, the major challenge is to cope with linear accelerations and magnetic disturbances. To that end, a data driven adaptive complementary filter for attitude and heading estimation is proposed. The accelerometers and magnetometers weights in each axis are adjusted according to optimized gain maps. The proposed approach is compared to commonly used algorithms using experimental data representing commonly used smartphone dynamics. Results show the contribution of the proposed approach to the attitude accuracy.