#openviriato grows out of its infancy

What has happened in the meantime

Four years have passed since our last report on the topic and the first invitation to participate in our #openviriato collaboration initiative. During this time, we have continued to work with our partners on the implementation of the programme. We initially focussed heavily on core investments, such as expanding the Algorithm Platform and building an interested network in the academic community. The Algorithm Platform was developed from the outset with the aim of providing the required data for a range of use cases, which was realised through an abstract data model of the interface ("AID model").

The AID model is a generic domain model. Due to this, development progress was initially slower, but this is now more than compensated for by the fact that work can be carried out in a resource-saving manner in the long term. Hardly any new functionality needs to be developed for each use case, as the basic functionality can be reused. The scope of the algorithm platform has now reached a state of maturity and the collaboration with top-class institutes and universities is beginning to bear fruit.

The resources that SMA can channel into this undertaking are relatively small compared to those of large companies in the railway industry. This forces us to be careful with our resources but also to promote trusting and long-term cooperation with partners in order to achieve practical results.

Successfully putting research into practice

Our approach has a number of benefits: firstly, the prototyping of an algorithm can be started quickly, because the algorithm platform accelerates algorithmic research because it provides the basic functionality required by researchers (data, views, services) and at the same time enables the user to evaluate the results of the algorithms on a state-of-the-art planning system.

In contrast to other approaches commonly used in research, the user works with their familiar timetabling system and not with prototypical GUIs, which brings the associated advantages: It is more error-resistant, the GUI is more stable, productive data can be reused without data transformations, exports or working on only partial extracts, and the results can be analysed in familiar modules. All of this helps to put research results into practice more quickly because, compared to the usual approach, an agile development process with short feedback cycles can be used to work on real-world cases.

Secondly, time and cost savings can also be achieved when taking algorithms into production, as only one GUI and its connection to the algorithm platform needs to be implemented instead of a completely new development of a data model and the provision of data. We never focus on automation for its own sake, but rather on the targeted automation of individual use cases of a clearly defined process.

Not everything automated: a realistic focus can take you further in the medium term

In recent years, there have been widespread attempts to create automated national service concepts. Such a project requires network-wide conflict resolution, which is known to be a problem that almost certainly cannot be solved in polynomial time. In the meantime, the initial euphoria has been followed by disillusionment and in some places this endeavour has already been abandoned and in others at least it is increasingly being met with realistic scepticism. SMA is focussing its automation efforts on process steps that typically occur in medium-term planning, i.e. 3 to 6 years before operation. In the past, this planning phase was rather neglected in most European countries, with the notable exception of Switzerland. In the past two to three years, the importance of medium-term planning has undergone a strong upturn, driven in part by the FTE/RNE initiative "Timetable Redesign" (TTR), but also by national projects such as the "medium-term concept of optimised capacity" (mKoK) in Germany.

Microscopic conflict resolution, but only locally!

In our work as consultants for the creation of large-scale timetable concepts, in collaboration with our customers, we have succeeded in introducing microscopic conflict detection as early as the medium-term planning stage and have since established this process. However, the manual resolution of these detected conflicts is still a very time-consuming and labour-intensive task in which creative thinking is of secondary importance. This is why the automation of conflict resolution in this phase of the process is currently a key focus of our #openviriato initiative. A first successful internship and student research project have already been completed while work on a further dissertation and a "Co-Pilot Conflict Resolution" prototype which brings the algorithm platform together with MoD, have now been started.

Freight path search is a requirement

The second focus of our research activities is the mesoscopic search for train paths. The use case for this in medium-term planning consists primarily of constructing multiple freight train paths within the available capacity. This is generally a time-consuming task with little creative potential. Viriato's mesoscopic model makes it possible to consider very large problems and solve them in a short computing time, unrivalled today by microscopic methods. In the future, we will also consider capacity constraints due to worksites and capacity objects in the context of TTR. Additionally, we will analyse to what extent it is possible to extend the path search from a mesoscopic conflict model to a microscopic one using the conflict resolution approach described above.

Robustness also benefits

Last but not least, our robustness module is an analysis tool that has already been and continues to be used successfully in consulting projects. The robustness tool is based on a train simulation that can be accessed via the algorithm platform and which has been kept open for extensions so that customers can integrate their own delay distributions and dispatchers. From our side, the main focus of future research will be on modelling the realistic behaviour of dispatchers and their simple implementation via the interface.

The journey continues

In recent years, through perseverance and hard work, our #openviriato strategy has laid the foundations for us to collaborate successfully with partners on future automation solutions for the railway industry. Our strategy aims to tackle the problems that can realistically be solved today in order to advance the railway quickly and achieve efficiency gains. The key to success for us has been that our interface is open to partners and different use cases, which we have achieved through an abstracted domain model and reusable services. The first successful use cases "robustness", " path search" and "co-pilot conflict resolution" as well as both completed and ongoing collaborations in research projects have validated our intentions.