Pavle Kecman

Rescheduling models for railway traffic management in large-scale networks

In the last decades of railway operations research, microscopic models have been intensively studied to support traffic operators in managing their dispatching areas. However, those models result in long computation times for large and highly utilized networks. The problem of controlling country-wide traffic is still open since the coordination of local areas is hard to tackle in short time and there are multiple interdependencies between trains across the whole network. This work is dedicated to the development of new macroscopic models that are able to incorporate traffic management decisions. Objective of this paper is to investigate how different levels of detail and number of operational constraints may affect the applicability of models for network-wide rescheduling in terms of quality of solutions and computation time. We present four different macroscopic models and test them on the Dutch national timetable. The macroscopic models are compared with a state-of-the-art microscopic model. Trade-off between computation time and solution quality is discussed on various disturbed traffic conditions.

Process mining of train describer event data and automatic conflict identification

Data records from train describer systems are a valuable source of information for analysing railway operations performance and assessing railway timetable quality. This paper presents a process mining tool based on event data records from the Dutch train describer system TROTS, including algorithms developed for the automatic identification of route conflicts with conflicting trains, arrival and departure times/delays at stations, and train paths on track section and blocking time level. Visualisations of the time-distance diagrams and blocking time diagrams support the analysis of incidents, track obstructions, disruptions, and structural errors in the timetable design.

Online Data-Driven Adaptive Prediction of Train Event Times

This paper presents a microscopic model for accurate prediction of train event times based on a timed event graph with dynamic arc weights. The process times in the model are dynamically obtained using processed historical track occupation data, thus reflecting all phenomena of railway traffic captured by the train describer systems and preprocessing tools. The graph structure of the model allows applying fast algorithms to compute prediction of event times even for large networks. The accuracy of predictions is increased by incorporating the effects of predicted route conflicts on train running times due to braking and reacceleration. Moreover, the train runs with process times that continuously deviate from their estimates in a certain pattern are detected, and downstream process times are adaptively adjusted to minimize the expected prediction error. The tool has been tested and validated in a real-time environment using train describer log files.

Predictive modelling of running and dwell times in railway traffic

Accurate estimation of running and dwell times is important for all levels of planning and control of railway traffic. The availability of historical track occupation data with a high degree of granularity inspired a data-driven approach for estimating these process times. In this paper we present and compare the accuracy of several approaches to model running and dwell times in railway traffic. Three global predictive model approaches are presented based on advanced statistical learning techniques: LTS robust linear regression, regression trees and random forests. Also local models are presented for a particular train line, station or block section, based on LTS robust linear regression with some refinements. The models are validated and compared using a test set independent from the training set. The applicability of the proposed data-driven approach for real-time applications is proved by the accuracy of the obtained estimates and the low computation times. Overall, the local models perform best both in accuracy and computation time.

Optimal allocation of buffer times to increase train schedule robustness

Reliability and punctuality of railway traffic are among the key performance indicators, which have a significant impact on user satisfaction. A way to improve the reliability and on-time performance in the timetable design stage is by improving the timetable robustness. In order to increase the robustness, most railway companies in Europe insert a fixed amount of buffer time between possibly conflicting events in order to reduce or prevent delay propagation if the first event occurs with a delay. However, this often causes an increase of capacity consumption which is a problem for heavily utilised lines. A sufficient amount of buffer time can therefore not be added between every two conflicting events. Thus, buffer times need to be allocated carefully to protect events with the highest priority. In this paper we consider the problem of increasing the robustness of a timetable by finding an optimal allocation of buffer times on a railway corridor. We model this resource allocation problem as a knapsack problem, where each candidate buffer time is treated as an object with the value (priority for buffer time assignment) determined according to the commercial and operational criteria, and size equal to its time duration. The validity of the presented approach is demonstrated on a case study from a busy mixed-traffic line in Sweden.

Adaptive, data-driven, online prediction of train event times

This paper presents a microscopic model for accurate prediction of train event times based on a timed event graph with dynamic arc weights. The process times in the model are obtained dynamically using processed historical track occupation data, thus reflecting all phenomena of railway traffic captured by the train describer systems and preprocessing tools. The graph structure of the model allows applying fast algorithms to compute prediction of event times even for large networks. Accuracy of predictions is increased by incorporating the effects of predicted route conflicts on train running times due to braking and re-acceleration. Moreover, the train runs with process times that continuously deviate from their estimates in a certain pattern are detected and downstream process times are adaptively adjusted to minimize the expected prediction error. The tool has been tested and validated in a real-time environment using train describer log files.