对绝大多数空间目标而言,由于观测条件和观测手段的限制,测轨数据稀疏且精度较低,传统轨道确定时一并解算的大气阻力系数精度稳定性很差.针对这一问题,提出一种基于随机森林的大气阻力系数预测模型.该模型主要利用某个目标的历史大气阻力系数、轨道数据、太阳地磁指数、大气密度等信息预测该目标未来一段时间内的大气阻力系数.仿真 GRACE(gravity recovery and climate experi-ment)A卫星2002年的测轨数据,进行多个时间段的轨道确定与预报模拟实验.结果表明,相比于传统方法,利用所提模型预测大气阻力系数,再将其应用于轨道确定和预报,7天轨道预报最大误差降幅可达60%,有效抑制了最大误差,为稀疏测轨数据条件下改善空间目标轨道确定与预报精度提供了一种技术途径.
Application of Random Forest in Determining Atmospheric Drag Coefficient of Low Resident Space
For most space objects,due to the limitations of observation conditions and observation means,tracking data is usually sparse and of low accuracy,and the accuracy of at-mospheric drag coefficients solved in the normal orbit determi-nation process,is very unstable.To solve this problem,we propose a predictive model to construct atmospheric drag coef-ficients based on random forest.It uses historical drag coeffi-cients,orbit data,the solar and geomagnetic activity indices,and the atmospheric density to predict the atmospheric drag co-efficient of the object for a future period of time.The orbit de-termination and prediction simulation experiments are carried out with simulation data of orbit measurement data of gravity recovery and climate experiment(GRACE)A satellite in 2002.The results show that the 7-day maximum orbit predic-tion errors can be reduced by 60%by using the proposed model to predict the drag coefficient and then applying it to or-bit determination and prediction,compared with the normal methods.The proposed model can suppress the maximum er-rors effectively,thus providing a technical way to improve the orbit determination and prediction accuracy for resident space objects under the sparse data condition.
atmospheric drag coefficientorbit predictionrandom forestsparse data
NETL
NSTL
万方数据
