Efficient Evaluation of Mars Entry Terminal State Based on Gaussian Process Regression

2018 2nd International Conference on Aerospace Technology, Communications and Energy Systems (ATCES 2018)

航空航天工程;无线电电子学;能源学

Gao, A.^1 ; Wang, G.Y.^2 ; Wu, S.S.^3 ; Song, T.^4

Beijing Institute of Technology, 5 South Zhongguancun Street, Beijing

100081, China^1

Beijing Institute of Technology, 5 South Zhon. Street, Beijing

100081, China^2

XD Group (XD) and General Electric (GE) Automation Co. Ltd., Xi'an, Shanxi province

710075, China^3

Shanghai Aerospace Control Technology Institute, Shanghai

201109, China^4

关键词: Atmospheric dynamics;    Ballistic coefficient;    Entry flight path angles;    Gaussian process regression;    Optimization parameter;    Optimization solvers;    Planetary exploration;    Trajectory optimization;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/449/1/012010/pdf
DOI  :  10.1088/1757-899X/449/1/012010

学科分类：航空航天科学

来源: IOP

【 摘 要 】

Trajectory optimization technology used for Mars entry is one of the key technologies for planetary exploration. Evaluation of the performance of the entry trajectory under conditions of complex atmospheric dynamics, various vehicular design parameters, and multiple constraints in the process of entry, are important issues pertaining to the design of trajectories. In this study, an efficient evaluation approach of the terminal state for Mars entry is proposed based on Gaussian process regression to evaluate the maximum terminal altitude for different entry velocities, terminal Mach numbers, and vehicular parameters. Additionally, the influences of entry flight-path angle, lift-drag ratio, and ballistic coefficient, on the maximum terminal altitude are analyzed. A genetic algorithm is used in the optimization solver to avoid local minima and to guarantee the data quality of the training samples used for Gaussian process regression. The mean function, kernel function, and hyperparameters are selected as the optimization parameters for Gaussian process regression to describe the correlation between samples, and the maximum terminal altitude prediction model is then established. Numerical simulations demonstrate that the proposed method can achieve the evaluation of more than 3000 group scenarios within tens of seconds with a mean relative error that is less than 4%.