Simulation modeling of pedestrian behavior in the presence of unmanned mobile robots

Abstract

Interactions between pedestrians and robots are becoming more commonplace. In public areas, for example, robots may be used for information dissemination, security, or patrol tasks. Based upon existing literature in the field of human-robot interaction, the ISAPT simulation system was revised to model individual pedestrian behavior in the presence of a mobile robot. Using an agent-based modeling approach, pedestrians are statistically assigned one of six reported behaviors when a robot is encountered: interact, watch, curious, ignore, cautious, and avoid. The modeling methods for incorporating these behaviors include modifying a pedestrian's existing agenda and/or their perception of the threat represented by the non-humanoid robot, while the pedestrian continues to make navigation decisions based on their overall utility function. This paper discusses the implementation of this capability and presents results on ISAPT's ability to reproduce the different behaviors reported in the literature. Data collected in a field study are used to further validate the system by comparing measures from observed behaviors to simulation output. Validation measures included lateral distance to robot and lateral path deviation. These results illustrate this approach is an effective means for adding this capability to microsimulation modeling systems.