Heterogeneous Speed Profiles in Discrete Models for Pedestrian Simulation

Abstract: Discrete pedestrian simulation models are viable alternatives to particle based approaches based on a continuous spatial representation. The effects of discretisation, however, also imply some difficulties in modelling certain phenomena that can be observed in reality. This paper focuses on the possibility to manage heterogeneity in the walking speed of the simulated population of pedestrians by modifying an existing multi-agent model extending the floor field approach. Whereas some discrete models allow pedestrians (or cars, when applied to traffic modelling) to move more than a single cell per time step, the present work proposes a maximum speed of one cell per step, but we model lower speeds by having pedestrians yielding their movement in some turns. Different classes of pedestrians are associated to different desired walking speeds and we define a stochastic mechanism ensuring that they maintain an average speed close to this threshold. In the paper we formally describe the model and we show the results of its application in benchmark scenarios. Finally, we also show how this approach can also support the definition of slopes and stairs as elements reducing the walking speed of pedestrians climbing them in a simulated scenario.