Greater Oslo Area Train Optimization

The railway is the greenest and most sustainable means of transportation, especially in urban areas. Growing urbanization is putting a tremendous pressure on regional transportation systems worldwide, including Norway. Indeed, the railway network that serves the greater Oslo metropolitan area has recently experienced increasing difficulties and delays. Building new infrastructure is very costly and difficult in densely populated areas. It is hard to assess what the impact of specific investment decisions will be, and it takes several years from when these decisions are taken to the actual finalization. On the other hand, there is plenty of unused capacity which may be exploited with better control of the railway traffic. Indeed, trains are currently dispatched manually, with very little support from digital tools. The GOTO main goal is to transfer the most recent advances in optimization and machine learning to railway traffic management. We will develop the methodological groundwork for an Optimization-based Traffic Management System that integrates state-of-the-art mathematical optimization algorithms and advanced forecasting techniques to tackle the complex scenarios of train dispatching that are found in Norway and the rest of Europe. Expected benefits include improved punctuality, reduced workload for dispatchers, and a more efficient utilization of the existing infrastructure. The outcome of this project will be a prototype tool that will be tested in the greater Oslo area. Dispatchers at the Oslo control center will be able to visualize the effects of each decision they make, up to few hours ahead. The system will automatically suggest them a set of optimized dispatching decisions based on the current train positions and preferences. All in real-time. In addition, we will investigate a novel methodology that exploits information gathered by the optimization algorithm to "learn" bottlenecks in the rail network and help develop the infrastructure. The most important result achieved during the first year of the project is the release of the first prototype in the second quarter of 2020, as planned. This tool can dispatch trains in an optimal way on a single line in the Greater Oslo Region. The software implements a mathematical optimization algorithm. The lines chosen for this first demonstration and evaluation were 1) Konsvingerbanen, running from Lillestrøm to Charlottenberg and 2) Gjøvikbanen, from Oslo to Gjøvik. The software takes as input the status of the railways and of the trains, and produces as output an optimal plan, namely routes and schedules of all trains running the line for the next 6 hours. The objective of the plan is to minimize delays, weighted by train priorities. The plan is recalculated every 10 seconds and displayed to dispatchers. To this end, we equipped the tool with an agile interface, which also allows dispatchers to modify decisions taken by the algorithm, such as departure times or meet and pass points, in order to evaluate the consequences of different choices on the traffic. During the second year, we focused on the development of mathematical models and methods for optimally dispatching trains in the large Oslo Central Station. The station contains 20 platforms and accommodates about 1000 trains per day, arriving from and departing to the 10 lines incident to Oslo S. The optimization algorithm returns an optimal or near-optimal solution every 30 seconds. The algorithm has been implemented in a prototype which allows dispatchers to inspect the suggested plans in an innovative GUI. This prototype was delivered officially on September 3 and is now available to dispatchers for the field-test campaign. In the final year of the project, we completed the study of the decomposition approach and delivered a third prototype of the dispatching decision support tool capable of simultaneously processing Kongsvingerbanen, Gjøvikbanen, Hovedbanen, Dovrebanen and Gardermobanen, plus an algorithm for Oslo central station. Also, the software system has been re-designed to be more flexible in treating the infrastructure model, so that we can more easily handle extensions and changes in the line and station models. The system has also been generally improved to achieve industrial software quality. In December 2022 the final version of the decision support tool was presented at Oslo control center during two meetings. Each meeting gathered a sub-group of the active dispatchers. The dispatchers are now allowed to test and play with a version of the prototype available on their workstation computers.

The main goal of the project was to develop the mathematics and the algorithms to implement a automatic tool able to make optimal dispatching decisions in real-time. The algorithms and the mathematics were successfully developed and the software prototype implemented and it is currently tested by dispatchers at Oslo control center. Some of the anticipated impacts (i.e. on punctuality, velocity, capacity and workload) can only be realized when the software will be actually rolled out and extensively utilized by dispatchers. However, the scheduling algorithm developed is state-of-the-art and tests on the real-life instances arising in greater Oslo area, show that the tool is able to produce good or optimal solutions for a large portion of network.

The GoTo main goal is to transfer the most recent advances in optimization and machine learning to railway traffic management. We will develop the methodological groundwork for an Optimization-based Traffic Management System (OTMS) that integrates state-of-the-art optimization algorithms and advanced forecasting techniques to tackle the complex scenarios of train dispatching that are found in Norway and the rest of Europe. Expected benefits include improved punctuality, reduced workload for dispatchers, and more efficient utilization of the existing infrastructure. The outcome of this project will be a prototype tool that will be tested in the greater Oslo area. Clearly, all developments for the greater Oslo area can be easily extended to any other railway region found in Norway. Dispatchers at the Oslo control center will be able to visualize the effects of each decision they make, up to few hours in the future. More importantly, the OTMS will automatically suggest them a set of optimized dispatching decisions based on the current train positions and preferences. All in real-time. In addition, we intend to investigate a novel methodology that exploits information gathered by the optimization algorithm of the OTMS with the purpose of identifying bottlenecks in the rail network. This will help steer future investment decisions in infrastructure enhancements. The project will last 3 years, and it will involve three main actors: SINTEF as the project owner and research partner, Bane NOR as the industrial partner and NSB as an end-user (train operator). SINTEF will contribute with its extensive expertise and outstanding research in optimization, machine learning, and train dispatching, in particular. Bane NOR, as the responsible for operating and developing the Norwegian railway network, will provide all the necessary knowledge and support for the prototyping and validation.