An improved CDKF algorithm based on RBF neural network for satellite attitude determination

In this paper, an improved Central divided-Difference Kalman Filter (CDKF) algorithm for the satellite that adopts three-axis magnetometer (TAM) and a fiber optic gyroscope (FOG) as attitude sensors is proposed. First, to overcome the disadvantages of Extended Kalman Filter (EKF), which is widely used as the attitude determination technique of satellites at the present, we present Central divided-Central Kalman Filtering (CDKF) algorithm applicable on the basis of the discretization kinematics equation of satellite, and applicable to attitude determination of this type of satellite. Then, considering the uncertainties and external disturbances on the system, we utilize RBF neural network to adjust the filtering gain online. Finally, a comparative simulation experiment is provided to show that our RBF-CDKF algorithm is able to achieve greater estimation performance than EKF and CDKF.