小推力转移燃料消耗估计的机器学习方法
|
摘要
深空探测任务设计初段往往需要求解复杂的全局优化问题。小推力轨迹的设计与优化问题精确求解较为复杂,求解速度较慢。由于计算能力与时间要求,不可能在全局优化的过程中对每一个方案都进行精确的小推力数值求解,所以在全局优化阶段需要对小推力转移进行快速准确地估计。采用机器学习的方法,对燃料最优小推力转移的燃料消耗进行了估计,其结果明显优于目前最为常用的Lambert估计方法。根据轨道描述方法的不同以及是否带有Lambert估计特征,采用不同的特征组合进行机器学习,分析结果发现带有Lambert估计特征的春分点轨道根数的特征组合为较好的机器学习特征组合。可为未来深空探测任务轨道设计提供参考。
It is often necessary to solve complex global optimization problems in the preliminary deep space mission design.The exact solution to the design and optimization of low-thrust trajectory is more difficult and time-consuming, because of the limitation of calculation ability and time,it's impossible to solve each low-thrust problem accurately using numerical methods in the global optimization process. In this paper, we propose a machine learning method to estimate the fuel consumption for fueloptimal low-thrust transfer. The results show the performance is better compared with the Lambert method which is commonly used at present. Different features are used for machine learning,and the major differences are different orbit description and whether Lambert estimation result is considered. The feature with equinoctial orbit elements and Lambert estimation is the best feature. It can provide reference for future orbit design of deep space exploration mission.
It is often necessary to solve complex global optimization problems in the preliminary deep space mission design.The exact solution to the design and optimization of low-thrust trajectory is more difficult and time-consuming, because of the limitation of calculation ability and time,it's impossible to solve each low-thrust problem accurately using numerical methods in the global optimization process. In this paper, we propose a machine learning method to estimate the fuel consumption for fueloptimal low-thrust transfer. The results show the performance is better compared with the Lambert method which is commonly used at present. Different features are used for machine learning,and the major differences are different orbit description and whether Lambert estimation result is considered. The feature with equinoctial orbit elements and Lambert estimation is the best feature. It can provide reference for future orbit design of deep space exploration mission.
引文
[1] GATTO G,CASALINO L. Fast evaluation and optimization of lowthrust transfers to multiple targets[J]. Journal of Guidance Control&Dynamics,2014,38(8):420-423.
[2] LI H Y,CHEN S Y,BAOYIN H X. J2-perturbed multi-target rendezvous optimization with low thrust[J]. Journal of Guidance,Control,and Dynamics,2018,41(3):802-808.
[3] AMPATZIS C, IZZO D. Machine learning techniques for approximation of objective functions in trajectory optimisation[C]//Artificial Intelligence in Space(IJCAI-09 Workshop). Pasadena,California in US:Artificial Intelligence in Space(IJCAI-09Workshop),2009.
[4] CASSIOLI A,DI LORENZO D,LOCATELLI M,et al. Machine learning for global optimization[J]. Computational Optimization&Applications,2012,51(1):279-303.
[5] HENNES D,IZZO D,LANDAU D. Fast approximators for optimal low-thrust hops between main belt asteroids[C]//Computational Intelligence(SSCI),2016 IEEE Symposium Series on 2016. Athens,Greece:IEEE,2016.
[6] CASALINO L,COLASURDO G. Problem description for the 7th global trajectory optimisation competition. GTOC portal[EB/OL].(2014)[2018-4-10]http://sophia. estec. esa. int/gtoc portal,2014.
[7] JIANG F H,BAOYIN H X,LI J F. Practical techniques for lowthrust trajectory optimization with homotopic approach[J]. Journal of Guidance,Control,and Dynamics,2012,35(1):245-258.
[8] PEDREGOSA F,VAROQUAUX G,GRAMFORT,A,et al. Scikitlearn:machine learning in Python[J]. Journal of Machine Learning Research,2011(12):2825-2830.
[2] LI H Y,CHEN S Y,BAOYIN H X. J2-perturbed multi-target rendezvous optimization with low thrust[J]. Journal of Guidance,Control,and Dynamics,2018,41(3):802-808.
[3] AMPATZIS C, IZZO D. Machine learning techniques for approximation of objective functions in trajectory optimisation[C]//Artificial Intelligence in Space(IJCAI-09 Workshop). Pasadena,California in US:Artificial Intelligence in Space(IJCAI-09Workshop),2009.
[4] CASSIOLI A,DI LORENZO D,LOCATELLI M,et al. Machine learning for global optimization[J]. Computational Optimization&Applications,2012,51(1):279-303.
[5] HENNES D,IZZO D,LANDAU D. Fast approximators for optimal low-thrust hops between main belt asteroids[C]//Computational Intelligence(SSCI),2016 IEEE Symposium Series on 2016. Athens,Greece:IEEE,2016.
[6] CASALINO L,COLASURDO G. Problem description for the 7th global trajectory optimisation competition. GTOC portal[EB/OL].(2014)[2018-4-10]http://sophia. estec. esa. int/gtoc portal,2014.
[7] JIANG F H,BAOYIN H X,LI J F. Practical techniques for lowthrust trajectory optimization with homotopic approach[J]. Journal of Guidance,Control,and Dynamics,2012,35(1):245-258.
[8] PEDREGOSA F,VAROQUAUX G,GRAMFORT,A,et al. Scikitlearn:machine learning in Python[J]. Journal of Machine Learning Research,2011(12):2825-2830.