The Dutch railway network is one of the most heavily utilized railway networks worldwide, while the trains are also highly interconnected and share conflicting routes. A broken signal, worn switch, or defect train, is sufficient to ruin the timetable and spread large delays all over the network.
The goal of this project is to develop new models and a new (predictive) control approach for anticipative management of railway networks. We aim at closing the loop between timetabling and train operations with a continuous feedback of train positions and field data to allow fast rescheduling of train paths in case of disturbances. Using predictions based on railway traffic models, an optimization procedure will find the most effective measures for regulating train traffic in case of disruptions like incidents, blockages, speed restrictions, and delays. This model predictive control (MPC) approach makes it possible to maintain an up-to-date feasible plan on a national network level with respect to the current state of the infrastructure (speed restrictions, blocked tracks), resources (rolling stock and crews), train delays, and expected conflicts.
The research contains three main ingredients:
- Monitoring: developing a methodology to actively monitor train positions, speeds, and infrastructure availability, and to determine up-to-date running time estimates;
- Predictive traffic model: developing a real-time railway traffic prediction model that can be updated continuously with the latest information and selected control decisions; and
- Model predictive controller: developing a model predictive control system that optimizes future control decisions by using predictions of the future behavior of the railway traffic.
The core of the MPC approach is the traffic model for which a switching max-plus linear system is proposed. A max-plus linear system is a discrete-event dynamic system characterized by synchronization constraints that excellently fits the modelling of large-scale railway networks. The conventional max-plus models assume a fixed structure of e.g. train orders and routings that possibly may not be maintained in perturbed operations. Switching max-plus models, however, can switch between different modes representing alternative decisions or circumstances.
Effective modelling procedures are to be developed to build typical control decisions into max-plus models, such as rerouting, reordering, and changing meeting or overtaking locations. In this project we will study the modelling and theory of switching max-plus linear systems with a focus on computational aspects and an embedding in a real-time MPC framework. Furthermore, we will develop a model predictive railway traffic management system to be used as a supervisory and intelligent decision support system for traffic controllers and signallers.
Utilization:
Railway traffic control currently involves a continuous monitoring and supervision of train positions, registration of the cause of train delays that are larger than 3 minutes, and application of rule-based local dispatching measures for many lines and stations in case of severe disturbances. Working in an anticipative manner is poorly supported and train traffic controllers are usually restricted to just solving problems reactively as they occur. This often results in unnecessarily long delays and a decreased reliability of the railway transport system.
In order to realize a change from reactive to proactive traffic control, algorithms and integrated software modules are to be developed to enable fast (semi-automatic) and optimal traffic management decisions on a network scale by traffic controllers and dispatchers. Advanced railway traffic models and control strategies are to be developed to optimize control decisions in real-time from a network-wide perspective. A key result in the research is the integration of the developed railway traffic management algorithms into a real-time simulation platform. Since our systems are based on formal theories our concept is generally applicable to other railway networks and related forms of mass transportation systems.
Proactive traffic management on the network level is a cost-efficient way to accommodate the expected growth of railway transport in the next decade without expensive infrastructure extensions. It improves train traffic punctuality and travel time reliability, increases utilization of existing and future railway infrastructure, passenger satisfaction and social welfare, and reduces energy consumption and costs due to congestion. This research thus provides a significant contribution to address the mobility problem.