Periodic variable navigation can realize the integration of positioning, attitude determination, timing and other functions. As a novel autonomous navigation method, autonomous management and autonomous operation of spacecraft are the main direction of great significance to reduce the burden of ground measurement and control, to improve the viability of spacecraft, While, the precision of the navigation is very critical for the use of periodic variable navigation for the long range spacecraft. So, we propose a big-data inspired precision improvement algorithm in this paper. Period variable star phase time measurement is used as observation information, and accomplished by the orbital dynamics equation of spacecraft motion. According to the gathering of the sampling data and the sensor data, a self-learning system is trained with the parameters from the nonlinear filtering methods. Based on the Unscented Kalman Filtering, an autonomous navigation algorithm of period variable star is established to realize the navigation and positioning of spacecraft. Under the measurement conditions of a single sampling interval of 0.01s and the measurement precision of period variable star phase observation time of 10^− 5s, It can be find that the navigation position determination precision can reach 400m, the speed determination precision can reach 0.3m/s, and the measurement precision can reach 10^− 5s with our proposed algorithm.

周期性变量导航可以实现定位，姿态确定，定时等功能的集成。作为一种新颖的自主导航方法，航天器的自主管理和自主运行是减轻地面测量和控制负担，提高航天器生存能力的重要方向。在远距离航天器中使用周期性变量导航。因此，本文提出了一种大数据启发的精度改进算法。周期可变星相时间测量用作观测信息，并由航天器运动的轨道动力学方程完成。根据采样数据和传感器数据的采集，用非线性滤波方法的参数训练一个自学习系统。基于无味卡尔曼滤波，建立了周期变星的自主导航算法，实现了航天器的导航定位。在单个采样间隔为0.01s的测量条件下，周期可变星相观测时间的测量精度为10s^{− 5} s，我们的算法发现导航位置确定精度可以达到400m，速度确定精度可以达到0.3m / s，测量精度可以达到10 ^{− 5} s。