摘要
列车精确停车是实现轨道交通自动控制系统的关键技术之一。传统的精确停车技术需要依赖于复杂的物理模型及昂贵的传感设备,且难以达到较高的精度。从数据本身出发,利用机器学习中高斯过程回归和Boosting回归算法对列车精确停车问题进行了研究,并与线性回归方法进行了比较,实验表明,机器学习的方法对于解决列车精确停车问题是行之有效的。其中以高斯过程回归的性能最优,而基于梯度的Boosting回归方法在缺乏先验知识的条件下达到接近高斯过程回归的性能,在实际应用中具有更大的灵活性和适应性。
Precise train stopping is one of the key technologies of automatic train control system.Traditional technologies of precise train stopping depend on complicated physical model and expensive sensor equipment,and it is hard to achieve high precision.The data themselves are utilized, applying Gaussian process regression and Boosting regression in the filed of machine learning, to study the problem of precise train stopping.The above methods are compared with linear regression.It is shown in the experiment that, the methods of machine learning are effective to the problem of precise train stopping.Gaussian process regression attains the best performance compared with the other methods.Gradient-based Boosting regression,with its performance approximating to that of Gaussian process regression in the lack of prior knowledge, demonstrates its flexibility and adaptability in practical applications.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第25期226-230,共5页
Computer Engineering and Applications
基金
国家自然科学基金No.60975044
轨道交通控制与安全国家重点实验室(北京交通大学)开放和自主课题基金No.RCS2008K007
No.RCS2009ZT004~~