| 1. |
- Crespo Materna, Arturo, et al.
(författare)
-
Use of Hybrid Methods for the Enhancement of Real-Time Railway Traffic Control (Dispatching)
- 2023
-
Konferensbidrag (refereegranskat)abstract
- The execution of scheduled railway operations is characterized by continuous monitoring and systematic adjustment of the existing schedule to the occurrence of stochastic events. The adjustment of the schedule can be referred to as the “Conflict Detection & Conflict Resolution” (CDCR) process. Caused by propagating conflicts between plan adjustments and the initially planned schedule, CDCR is a highly complex process. Due to complexity, a series of decision-support tools mostly relying on heuristic methods have been developed to assist dispatchers in real-time. This article aims to identify strategic enhancement potentials for improving existing schedule adjustment approaches by integrating different methods (e.g., machine learning methods). A decomposition method is utilized to identify the processes during schedule adjustment that could benefit from applying hybrid methodologies, resulting in a much more efficient and effective search space exploration. At the outset the processes of generating a set of conflict resolution alternatives and selecting the best-fitting alternative to the actual operating situation have been early identified as potential processes that would benefit from incorporating hybrid methods (e.g., machine learning and heuristic methods). This study utilizes an actual decision-support tool applied within a real scenario to derive concrete evidence regarding the extent to which hybrid methods can be integrated and used to solve complex problems within the real-time adjustment of railway schedules by means of their actual implementation in an existing process. The knowledge and experience gained from the experimental research, acting as a proof of concept, are then translated into general guidelines for further use in improving existing approaches used in decision-support tools for the CDCR.
|
|
| 2. |
- Högdahl, Johan, Doctor, 1989-, et al.
(författare)
-
Maximizing railway punctuality : A microsimulation evaluation of robust timetabling methods
- 2023
-
Konferensbidrag (refereegranskat)abstract
- Punctuality is commonly recognized as one of the most important quality indicators for passenger traffic. Despite this, surprisingly few methods for explicitly maximizing punctuality by optimizing the timetable exists in the literature. We study how late-stage adjustments during the capacity allocation can improve punctuality of the traffic. In this paper, we therefore extend a combined simulation-optimization method so it can be used to explicitly maximize the predicted punctuality of a given nonperiodic timetable on a double-track line. The method is evaluated in two microsimulation experiments in the southbound direction of the Swedish Western Main Line using Railsys. We compare the method in simulation with our previous method for minimizing total disutility, two methods from the scientific literature (light robustness, and robustness in critical points) and two naïve strategies. The methods’ effectiveness is assessed in a detailed statistical analysis considering end-station punctuality, total punctuality, and the robustness measure total disutility. Only light robustness results in timetables that in simulation performs better or equally well as the given timetable (based on the national timetable) with respect to all performance measures and evaluated scenarios. The method for maximizing punctuality performs best with respect to total punctuality.
|
|
| 3. |
- Högdahl, Johan, Doctor, 1989-, et al.
(författare)
-
Reinforcement Learning Based Robust Railway Timetabling to Resolve Robustness Vulnerabilities
- 2023
-
Konferensbidrag (refereegranskat)abstract
- Railway timetables have an important role in efficient and punctual railway operations. In particular, the robustness of the timetable has a direct impact on the traffic's punctuality. To evaluate the robustness of a timetable, simulation is commonly used. A simulation study may indicate that some trains are too sensitive against minor delays, which may lead to that they fall out of their planned channel of operations (defined by their surrounding trains). We define this as robustness vulnerabilities of the timetable. The work explores reinforcement learning (RL) as a method to resolve timetable robustness vulnerabilities. We formulate a RL-based model for the robust railway timetabling problem and will explore different RL algorithms and compare with timetables generated using optimization-based methods from our previous work [1, 2]. The models are evaluated using microscopic RailSys simulation for the traffic in the westbound direction of the Swedish Western Main Line. The results are expected to provide better support for robust railway timetabling in practice.
|
|
