Simulation of train driver behaviour using a data-driven agent

Abstract

The Dutch railway infrastructure is getting crowded with over 390 million passengers last year. Several innovations are created to improve the capacity by improving the infrastructure as well as the driving behaviour of train drivers. Within the latter, pilots are performed to gather feedback. However, feedback gathering is done on a per train basis and disregards the effect on a larger scale. Within the railway industry, trains are being monitored with multiple systems for different purposes. This thesis compares several monitoring systems and from some a model representing a train driver is created. The behaviour of multiple train drivers is simulated using a multi-agent system. An agent is used to simulate the behaviour of a train driver on a larger scale. The effects of the simulation are measurable at certain points, and the entire route is compared to trains without innovation giving a relative gain, hence the benefit of the innovation.