The 2018 FIFA World Cup is a major sporting event which takes place in 11 Russian cities. For the operational transport masterplan of this event a dynamic decision support system was built, which has a global MATSim model as a key element. The system created allows for constant update of the input data such as the list of qualified national teams, game schedule, the behavior model for guests and participants, global air traffic and railway schedule, etc. The input files for the MATSim scenario can be thus automatically generated based on the updated variables, which are stored in the database. The model then calculates an updated optimal schedule for charter and additional flights as well as capacity bottlenecks of the railway stations, airport terminals and runways of the host cities.
The scenario covers more than 6 million intercity and international trips during the period of more than 40 days within one simulation. A new behavior model has been implemented which covers the behavior of international spectators of a large event and is based on Markov chains.
The total amount of scheduled flights and railway departures are obtained from open web databases and simulated in the scenario along with car traffic.
In order to accomplish specific tasks of the transportation masterplan, various modifications and extension were created for the MATSim platform, which showcases its great flexibility and versatility. For instance, the existing RAPTOR routing module was used for transit simulation and was partially rewritten in order to simulate the intercity transit ticket buyout and to allow for computationally efficient complex routing of agents with a large number of transfers.
Another contribution of this project is the implementation of evolutionary algorithms for creating and selecting effective mutated additional regular flights to cover the additional global travel demand generated by the event.
A web-based service with version control was designed for visualization and expert analysis of the results of different scenarios, as well as for modification of the model parameters and variables. The results include simulated hourly passenger flows on the transport knots in all directions as well as the number of takeoff and landing operations.