Accounting for the Indecision of Pedestrians
Chris Lee introduces the research of Alessandro Corbetta, from the Technical University of Eindhoven, into the difficulties of creating reliable models for pedestrian behavior.
Unfortunately, pedestrian models are not very well tested against data. Most experiments involve paying university students to walk along corridors and through doors under highly artificial conditions. In part, this is because it has been very difficult to obtain data from natural settings, where you need to track individual pedestrians as they walk through some area of interest.
Corbetta set [Kinect] cameras up in two locations: the main thoroughfare at Eindhoven train station and a link corridor between one of the university buildings and the nearest cafeteria. From there, he recorded data for a year.
Corbetta's findings after crunching the quarter of a million trajectories he discovered during the experiment revealed some of the behavior that makes pedestrian modeling so difficult. Namely, that some people change their minds en route, and turn around. The article goes into more detail about how Corbetta accounted for their indecision for the ongoing project of building reliable pedestrian models.