基于CART算法的卫星星座原始构型选择策略

Strategy of selecting original configuration for satellite constellation using CART algorithm

针对现有的3种最具代表性的卫星星座设计方法的缺点,提出了层次递进的卫星星座设计方法.该方法采用基于CART算法的卫星星座原始构型选择策略,可将星座构型的求解空间缩小至少53.3%.利用CART6.0软件构建了选择卫星星座原始构型子集的CART树,运行速度达到7.9s,分类精度达到97.15%,10阶交叉验证的相对成本为0.033.结果表明:采用这种策略选择卫星星座原始构型还具有可扩展性强、准确性高和易于实现的优点,为卫星星座构型模块化设计建立了良好的基础.

Stepwise forward method of satellite constellation design was presented, which inherited the advantages of three most representative satellite constellation designs and could greatly reduce the solution space of satellite constellation design. The key to reduce the solution space was the strategy for original configurations selection of satellite constellation based on the CART （classification and regression tree） algorithm. With this strategy, the solution space was reduced by at least 53.3 %. A CART tree was constructed to select original configuration subset by CART software. The experiment shows the CART construction speed has reached 7.9 s, the classification accuracy is 97.15%, and the relative cost of 10-fold cross-validation is 0. 033. The results also show that the strategy offers many practical advantages, such as high scalability, high accuracy, and the easiness to implement, which establish a good foundation for the modular design of satellite constellation configuration.