vai al contenuto della pagina vai al menu di navigazione
 

An effective potential for the pedestrians dynamics in a socially interacting group

by Dr. Francesco Zanlungo (Advanced Telecommunications Research Institute International-Intelligent Robotics and Communication Laboratories - Kyoto Giappone)

Jul 22, 2014 from 11:30 AM to 01:00 PM

Where Aula Teorici, via Irnerio 46

Contact Name

Add to your calendar

Abstract:

In a recent work we introduced a Non-Newtonian potential, i.e. generating forces not obeying the third law of dynamics, to describe the spatial dynamics between pairs of socially interacting pedestrians in a walking group. We showed that under scarce density conditions, i.e. under the assumption that collision avoidance is not the leading dynamical term and can be modelled as a random noise if observed under a long enough time scale, our potential may be used to derive the statistical properties of the spatial distribution of pedestrians in small groups, along with the average velocities of the groups. By comparing to a data set of pedestrian group trajectories under low density conditions (rho=0.03 p/m2) we verified that the proposed model describes in a quantitatively accurate way the spatial structures and the velocity of 2 and 3 people groups (along with those of the, usually unstable and short living, fully interacting 4 and 5 people groups). We apply the same method to a new data-set, obtained by automatically tracking with 3D sensors, and video recording for subsequent visual analysis and recognition of groups by a human coder, the movement of pedestrians in a indoor facility (the ATC center in Osaka). By performing this analysis, we test the validity of our model under higher density conditions, and more generally study the effect of growing density on the pedestrian group structure. To describe in a qualitative way the variation of the group structure with changing density, we introduce an ‘effective potential’ term that assesses the average effect of the external environment on the group dynamics.