基于非参数贝叶斯模型的列车卫星定位方法

基于卫星导航的列车测速定位是当前新型列车控制技术领域的研究热点。针对现场运行环境中卫星定位观测特性的不确定性对动态定位解算性能的不利影响,在卫星定位估计解算过程中引入基于Dirichlet过程的非参数贝叶斯模型,在导航卫星观测特性不定且存在性能退化风险的条件下,提出了基于Dirichlet过程混合模型驱动的高斯分布混合更新策略,给出了多滤波器动态联合状态估计计算方案。分别采用仿真数据及现场实测数据进行计算的结果表明,在常规定位估计解算中引入Dirichlet过程混合模型,能够对未知不确定的导航卫星观测特性实现有效适应,提升定位性能对观测条件的容忍能力,降低单一观测分布假设的模型失配和性能劣化风险。

Satellite-based Train Positioning Method Based on Non-parametric Bayesian Model

Satellite-based train positioning has become a research focus in the field of new generation train control technologies and systems.In this paper,considering the negative effects of uncertainties in satellite positioning observation characteristics in field operation environment on dynamic positioning computation performance,the Dirichlet process-based non-parametric Bayesian model was introduced into state estimation in the satellite navigation calculation process.To cope with uncertain observation conditions and performance degradation risks,a Gaussian mixture update strategy driven by the Dirichlet Process Model(DPM)was presented.A dynamic weighted state estimation method was given for parallel sub-filters.The results of simulation data and field experiments demonstrate the effectiveness of the DPM-based state estimation method in dealing with unknown and uncertain satellite measuring characteristics.This method is capable of enhancing the tolerance of positioning performance against different observation conditions and reducing the risk of model mismatch and performance degradation under single observation distribution assumption.It allows remarkable potentials in the application of future satellite-based train control systems and other related implementations.