Calibrating a General Pedestrian Stream Simulation Model According to a Specific Real Life Scenario of a German Railway Station

Pedestrian stream simulations are used to mitigate risks for people in buildings and at public events. They allow gaining virtual experience for situations where it is impossible to gather real experience. At some point even short term predictions may become possible. However, simulations remain useless if the simulation tool does not reliably reproduce the true phenomena. Hence, one of the biggest challenges in the applied research field of crowd motion is the validation of the proposed models. Up to now very little data from real-life scenarios has been available and calibration attempts have mostly relied on literature values or, at best, laboratory measurements. Instead, we propose to extract key data from live video records as a part of a methodological approach that makes possible to calibrate a simulation tool against video data. We re-enact a real-life scenario observed at a major German railway station with a benchmark simulator and compare the results to the video observations to give a proof of concept for our approach.