Online Trajectory Planning Method For Small Asteroid Based On Neural Network

Since the 20 th century,asteroid exploration missions have been on the rise.Major countries and organizations in the world have carried out or planned asteroid exploration missions.The China National Space Administration has also put forward China’s asteroid exploration plan in the coming decades.Asteroid exploration mission has become an important form of international competition,representing a country’s social development level and comprehensive national strength.At the same time,asteroid exploration mission is also conducive to better understanding of the formation and evolution of the solar system,and to better understanding of nature.In this paper,Eros433 asteroid was taken as the target asteroid,a neural network model of the asteroid gravitational field was established to optimize the landing trajectory of the probe,and a fast planning method of the landing trajectory and an autonomous landing control method were proposed.First,the polyhedron model of the asteroid’s gravitational field is established in this chapter,and then the neural network model of the gravitational field is established based on BP neural network.In addition,the mechanical model and three-dimensional dynamic model of asteroid disturbance are established.After that,this paper introduces the direct method and the indirect method of the trajectory optimization method in detail,and USES the pseudo-spectral-homotopy hybrid method to solve the trajectory optimization problem under the strong nonlinear gravitational field,and generates the data set corresponding to the initial state value and the initial costate value.Then,BP neural network and ELM neural network are respectively used to fit the corresponding relationship between the initial state value and the initial costata value,and compared the advantages and disadvantages of each other.Finally,based on the previous work,the autonomous control method of asteroid soft landing is proposed.In other words,the neural network corresponding to different landing time and costate fitting is generated.During the landing process of the probe,the costate value(control quantity)of the probe is modified at intervals to control the probe to reach the target landing point.