Prediction in social path following

Abstract

Path following in games mostly focuses on avoiding collisions with dynamic physical objects that appear along a chosen path to a given destination. Some work also attempts to humanize the abstract path returned by a path finding algorithm through methods like smoothing. Games typically do not consider social factors during path following, even though many depict social environments. Social path following considers the social environment in particular, carving a trajectory that reflects awareness of other human beings and their social activities. This includes awareness of territories that have social significance but no concrete physical form, such as the space between those having a conversation. This paper describes work that extends a state-of-the-art predictive method for path following with social awareness, predicting and avoiding social collisions. The work builds on a platform for social simulation, which already models social territoriality and gaze behavior. The results appear promising and highlight the importance of perceiving and dealing with the social space along with the physical one.