Anticipation manifests itself at various levels in pedestrian dynamics. This paper aims to clarify its impact on the modeling of local navigation, by shedding light on two facets of anticipation over the near future, one related to the others’ motion and the other one related to the agent’s own prospective motion. We start by observing that in simple (‘reactive’) models anticipation is completely missing, whereas in velocity-based models, it generally rests on linear extrapolation, with the use of e.g. time-to-collision metrics. Focusing on a crossing scenario, in which a pedestrian has to cross a static crowd, we pinpoint some of the ensuing limitations. We conclude by presenting preliminary data on a modeling framework that attempts to bridge the gap between the above mainstream models and full blown game theory.
Raulin-Foissac et al. (Fri,) studied this question.