No abstract

No abstract

To improve the situation for crowded commuters in Japan, it is important to plan a train schedule that considers passenger behavior, such as their choice of trains and the transfer stations used to reach their destinations. However, it is difficult to directly measure such detailed behavior using the present infrastructures, with which we can only get OD (Origin-Destination) data from the automatic ticket gates. The obtained OD data only consists of the number of passengers for each origin-destination and the time each passenger passes through the gates. In this article, to contribute to the planning phase of a new train schedule, the authors propose a method for estimating railway passenger flow using OD data. This paper firstly points out that the problems of estimating passenger flow can be boiled down to a shortest path problem of graph theory by assuming a certain passenger behavior model. By representing train operations in a graph structure, we can assume that a passenger will use the minimum cost path to his/her destination. This paper secondly proposes a method for conducting fast searches of the graph structure. The method uses the fact that railways operate on a time schedule. This method can estimate passenger flow fast enough so as to apply it to a practical train schedule planning support system. Lastly, the authors show the results of applying the passenger flow estimation system to a railway in an urban area in Japan.

We propose a general prediction method based on the efficient computation and online update of the Singular Value Decomposition (SVD) of historical data. The SVD is fundamental to many data modeling algorithms, but the traditional methods for computing it require large computational costs. By adopting a fast sequential SVD updating scheme, the tasks of prediction, imputation of missing values, and model updating can be performed very quickly. In this paper, an application of our method to route travel time prediction is described. Using real travel time data from short sections (links) on expressway, we evaluated prediction performance of travel time on longer section (route) including the links. Experimental comparisons with several statistical machine learning methods suggest that our linear prediction method can achieve similar prediction performance (prediction error) to other nonlinear methods at less computaional cost.

Automatic train control plays a key role in improving the efficiency and safety of train movements, as well as the riding comfort of passengers. In Japan, train control systems have been successfully implemented since 1980s. These systems have been required to obtain train position and speed as accurately as possible. This has mostly depended on axle generators and transponders. More specifically, the axle generators measure the speed and the moving distance from the reference points specified by the transponders. However, train control systems using these devices still fail to get correct train position, due to skidding or slipping, until passing over reference points. This paper focuses on train automatic stop control (TASC), and presents a new TASC system using a commercial range sensor instead of transponders so that a train implementing the system can detect its position continuously.