Towards more behaviours in crowd simulation

Abstract

While collision avoidance has been the most active topic in pedestrian simulation, the modelling of other kinds of behaviours appears to be essential for better realism. Thus higher cognitive levels of perception and behaviour improve simulation quality. Furthermore, giving an agent the possibility to choose the nature of its interactions with the others can not only improve simulation realism but also bring heterogeneity in the simulated population because each agent individually perceives the situation according to its own characteristics. In this paper, we aim at providing the pedestrian agent the ability to obtain an individual representation of the environment that allows him to adapt its behaviour according to the situation. We base our work on the analysis and interpretation of the environment, which makes the agent decide the behaviour it is going to adopt. We focus on two kinds of behaviours, following and group avoidance behaviours, and on their integration in classical avoidance simulations. We integrate recent works about following behaviour and propose to model interactions directly with groups of people instead of individuals. We aim at providing perception rules totally independent from the collision avoidance model used in the simulation. Because of the improved perception process, we observe emerging speed waves, group behaviours and lane formation in our simulations. Our results demonstrate the interest of modelling such behaviours to obtain more realistic simulations and show that specific patterns and collective behaviours emerge when using several types of behaviours in simulations.