Abstract: | When the number of formation satellites is large, the formation reconfiguration planning brings the huge computational cost considering collision avoidance. In order to reduce the computational overhead and improve the optimization efficiency, based on the CW equation and the dual pulse orbit maneuver strategy, a variety of surrogate models that can quickly predict the shortest distance in the formation satellite reconfiguration process are established. Based on three training sets of different sizes, the model accuracy and efficiency are compared. The result shows that Kriging (KRG) model has the highest accuracy among various surrogate models. Moreover, with the increase of training, the performance of Kriging and artificial neural network (ANN) models has been significantly improved, and the accuracy of the model is guaranteed. It is also found that although the prediction time of KRG model is higher than that of other surrogate models, its time-consuming is still very short compared with the real model, so it can be used to improve the efficiency of satellite formation reconfiguration trajectory optimization considering collision avoidance constraints. |