A framework of gravity field online modeling and trajectory optimization in asteroid soft-landing mission scenarios

Focused on asteroid soft-landing mission scenario, this paper investigates the methods of online gravity field modeling and trajectory optimization. Since the gravity field near asteroids and its surface topography are irregular, accurate asteroid soft-landing can be a challenging problem to solve, especially in the case of limited gravity data. With an actual asteroid soft-landing mission in mind, this study constructs a dataset based on the gravity field data collected in the limited area through which a soft-landing probe passed, and conducts online modeling of the asteroid inhomogeneous gravity field at the same time, using a transfer learning deep neural network (TL-DNN). To achieve fuel optimization and an accurate landing, a fast trajectory optimization method based on Bezier shape approach is implemented to achieve fast online robust re-trajectory optimization. Numerical simulation results show that the adopted TL-DNN can quickly reconstruct the gravitational field near the asteroid online, based only on the limited gravity data collected by the probe. Compared with the situation without online gravity field reconstruction and multiple online trajectory optimization, the proposed online gravity field modeling and re-trajectory optimization method can reduce the final position and velocity error between the multi-planned ideal optimal trajectory and the actual trajectory from 2396.367 m and 3.870 m/s to 14.152 m and 0.075 m/s, respectively. In other words, the soft-landing accuracy is improved by about two orders of magnitude.