Validating Social Force based Models with Comprehensive Real World Motion Data

• DOI: 10.1016/j.trpro.2014.09.080
• Publication Date: 2014-10-02T15:30:45Z
• Authors: Stefan Seer; Christian Rudloff; Thomas Matyus; Norbert Brändle
• Keywords: Social Force Model; Data set; Tracking (education)

Over the last years multiple variations of Social Force model have been proposed. While most available force-based models are calibrated on observed human movement data, validation for investigating characteristics, e.g. variance in parameter values, is still sparse. We present a novel methodology validating based which investigates reproducibility behavior individual trajectory level with real-world data. Our approach estimates values and their distribution non-linear regression where resulting variances represent model's validity. demonstrate our comprehensive (235 pedestrians) highly accurate (within few centimeters) set trajectories obtained from pedestrian traffic bidirectional flow using an automatic people tracking Kinect sensors. validate Helbing Molnár (1995), Johansson (2009) Rudloff et al. (2011).