Train dispatcher

For the past ten years, SINTEF Optimization has been leading the way in the development of AI technologies that will reshape the future of railways. We have collaborated with several industrial partners and academic organizations to optimize railway operations by combining state-of-the-art mathematical modelling and machine learning techniques. This project represents the first step towards the commercialization of an intelligent and automatic decision support system for real-time train operations, which makes continuous optimal adjustments of train schedules in response to traffic disruptions. Compared to existing manual procedures, this system can significantly increase network utilization while improving punctuality, with estimated social savings of millions of dollars per year for Norway alone. The main challenge to achieving optimal railway operations is the computational complexity arising from the so-called combinatorial nature of the underlying decision process. Consider, for example, the task of finding the right sequence of train operations within a station to satisfy all constraints while minimizing delays. More importantly, consider that decisions taken for a certain station can later affect many other stations, creating a cascading effect that leads to an exponential number of possible solutions. Like in a chess game, one cannot possibly hope to explore all solutions to find the best one. Instead, we combine different AI methods to explore only the most promising ones. This allows us to find close to optimal solutions in real-time (i.e., within a few seconds). Within this project, we work together with SINTEF TTO to identify business opportunities and go-to-market strategies. We are also establishing a collaboration with a world-leading provider of train management systems to validate our technology on real-life scenarios.