| Contents |
| Crowd models are powerful tools aiming at exploring complex dynamics of pedestrians walking in a crowd. In order for these models to have some predictive power, and hence being of use in civil engineering (e.g.~when pedestrian evacuation is concerned), the parameters on which they depend must be estimated.
It is not hard to imagine, though, that the behavior of pedestrians, even in simple scenarios, features a wide heterogeneity, determined by the so called \textit{inter-subject} and \textit{intra-subject} variabilities. Such heterogeneities make the task of estimating free-parameters by means of experimental data (e.g.~by detailed pedestrian trajectories) both \textit{unclear} and \textit{unconventional}.
We present a Bayesian probabilistic framework, which, on the basis of given experimental data, estimates the values and quantifies uncertainties (in the form of probability density functions) in the parameters of chosen models. In this framework, we also introduce a fitness measure for the models to classify several model structures (forces) according to their fitness to the experimental data, preparing the stage for a more general model-selection and validation strategy inspired by probabilistic data analysis. The results are based on a large amount of experimental pedestrian data obtained after a long time measurement at Eindhoven University of Technology. |
|