列车轨道曲线拟合算法改进研究

Research on Improvement of Train Curve Fitting Algorithm

为了获得高效、高精度的轨道曲线,为铁路运营维护提供准确的空间分析手段,实现列车定位和运行控制。在论文中利用历史定位数据的误差特性,提出在线学习的逻辑回归改进算法模型,通过最速下降法优化加快收敛,拟合多条轨道曲线数据。验证在交叉轨道、平行轨道等复杂情况的可行性,并且分析在不同Lambda(学习率)情况下对多条轨道拟合准确度的影响。仿真结果表明,论文逻辑回归拟合改进模型算法比逻辑回归拟合算法准确度高;在合理选择Lambda(学习率),能够在同一时间拟合多条轨道曲线数据并保证拟合度准确性达到90%以上。

In order to obtain high efficiency and high precision track curve,it provides accurate spatial analysis means for railway operation and maintenance to realize train positioning and operation control. In this paper,the error characteristics of historical positioning data are used to propose an improved logistic regression algorithm model for online learning. It can accelerate the convergence by the steepest descent method and fit multiple orbital curve data. It verifies the feasibility of complex conditions such as cross track,parallel orbit,and analyzes the influence of multiple orbital fitting accuracy under different Lambda(learning rate). In this paper,the improved regression model is more accurate than the logic regression fitting algorithm. In the reasonable selection of Lambda(learning rate),it can fit multiple orbital curve data at the same time and ensure the accuracy of fitting degree is over 90%.