The railway industry may not be the most eager to adopt new technological solutions, but it has always been a great source of interesting optimization problems. Many of these problems are extremely challenging to solve, partly because one needs to model realistic and safe train operations, and partly because of the sheer complexity of the railway infrastructure. But optimization-based decision support tools are needed now more than ever, since train traffic is expected to significantly increase, and new tracks are expensive to build.

In this context, SINTEF has been a pioneer in developing innovative optimization solutions for train scheduling problems. Scheduling in railways typically concerns three levels of planning: dispatching (optimized in real-time), tactical timetabling (a few days to one year in advance), and strategical timetabling (a few years in advance). Each of these problems required to introduce new modelling and algorithmic techniques, typically based on mathematical programming. A partnership with Gurobi allowed their customers to be sure that the optimization problems were always solved as fast as possible.

When dealing with industrial applications, it is typically hard to compare computational efficiency of different solutions from the scientific literature, because each manuscript considers a very realistic but slightly different optimization problem. And the literature on train scheduling applications is no exception. For this reason, they decided to collect instances from several countries, build a representative set of well-defined benchmark instances, and invite teams from all over the world to compete in a train dispatching competition.

In this webinar, we present the theoretical framework behind their train scheduling solutions, and we discuss some of its practical applications. Plus, we introduce the competition and explain its challenges.