An Estimation Algorithm of Attitude and Heading Under Homogenous Field Based on Improved Gradient Descent Method

With the development of MEMS (Micro-electromechanical Systems) manufacturing technology, MEMS inertial sensors have been widely applied in military industry and civil industry due to its advantages of low cost, low power consumption and small size. Although MIMU (MEMS-Inertial Measurement Unit) cannot meet the requirements of pure inertial navigation because of its low precision, it can be qualified for some specific navigation tasks by combining external data such as GNSS (Global Navigation Satellite System) and magnetic information with the fusion algorithms. MIMU is usually taken as the core sensor of AHRS (Attitude and Heading Reference System), meanwhile triaxial magnetometer is used to assist measuring attitude and heading with the gradient descent method. However, in the common gradient descent attitude estimation algorithm, the update step is unit size or just related to angular velocity. Hence, the estimated value of attitude converges slowly when the platform is stationary and the estimation result is unstable under the large angular velocity condition. In order to solve these problems, an estimation algorithm of attitude and heading based on improved gradient descent method is proposed in this paper. An inexact search method is adopted to obtain the optimal step length in each update, that improves the speed and stability of attitude estimation. The simulation results show that the estimation attitude of the improved algorithm can quickly converge to an accurate result in the condition of large initial error and the estimation precision is higher than conventional algorithm.