Spacecraft undergoes several operation modes during its lifetime such as de-tumbling mode, attitude acquisition mode, stand-by mode, and high accuracy mode. Some of these modes are characterized by high nonlinearity while some are characterized by linearity. Thus, a high performance nonlinear discrete controller (HPNDC) is utilized. The controller behaves like a nonlinear attitude control algorithm when dealing with maneuvers with large spacecraft attitude angles, and behaves like a linear PD controller for maneuvers with small angles. The gains of the controller are calculated to account for the discretization process. Controller stability is proven using Lyapunov second method. The controller can effectively deal with the attitude dynamics and kinematics represented by a six degrees of freedom system. In order to avoid singularities in the attitude representation, the rotation is expressed by a quaternion. A quaternion error vector is utilized to express the difference between the spacecraft direction and the target direction. An adaptive neuro-fuzzy inference system (ANFIS) controller is developed in order to mimic the performance of HPNDC. The Fuzzy Inference System is Sugeno-type. In order to model the training data set, a combination of back-propagation gradient descent methods and the least-squares algorithms is used. A weighted average defuzzification is used to obtain the output. Training ratio is 0.85, and the rest of training data is utilized to test the training data overfitting. Disturbance torques acting on the spacecraft are mainly due to gravity gradient torques which are considered as the dominant effects for low earth orbit spacecraft. The high initial angular velocities and attitude angles are nullified by the ANFIS controller. The proposed algorithm can bring the spacecraft from the de-tumbling mode to the high accuracy mode within two orbits.
Read full abstract