Freight Train Optimization and Simulation

Train scheduling has already received a lot of attention, whether for passenger or freight trains. While the volume of goods transport has increased over the years, extensions of railway systems are very rare because they represent major investments for railway companies or governments. Accordingly, the railway companies are often operating freight trains in a system that is close to saturation. It follows that a very effective planning and optimization of the rail network is needed. 
\n
\nWhile passenger train schedules are relatively static and cyclic, and can be planned months ahead, freight train schedules are designed with a much shorter planning time period, sometimes even one day or few hours before train departures. Moreover, passenger train schedules must obey some strict time window constraints as trains must arrive and depart from stations in order for passengers to get off/on the trains according to the posted schedule. 
\nOn the opposite, the schedule of the freight trains may vary according to the train lengths or loads, i.e., freight trains have a much greater variability in their average speed. Lastly, the track configuration of the freight trains does not have a dedicated direction as it is often the case for passenger trains. For all those reasons, the scheduling of freight trains is more complex than for passenger trains.
\n 
\nIn this thesis, we propose a new dynamic row/column management algorithm for the schedule of freight trains in a single/double track railway mesh network system. While many works have already been devoted to train scheduling, previously published optimization models all suffer from scalability issues. Moreover, very few of them take into account the number of alternate tracks in the railway stations or in the sidings for train meets, as well as the delay incurred by trains that take sidings. We propose a non time-indexed model, which takes into account such constraints, and we design an original solution scheme with iterative additions/removals of constraints/variables in order to remain with a manageable sized mixed integer linear program, while still ensuring convergence to an optimal solution. Numerical results are presented on several data instances of CPR (Canada Pacific Railway) on the Vancouver-Calgary corridor, one of the busiest corridors in their railway system.
\n
\nIn addition, we developed a simulation tool within the Arena framework, for the scheduling of freight trains. Comparisons of the simulation and optimization tools are made, together with a review of the pros and cons of simulation against optimization tools.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.