Competence in pedestrian social navigation requires a robot to exhibit many strengths, from perceiving the intentions of others through social signals to acting clearly to convey intent. It is made more difficult by the presence of many individual people with their own agendas as well as by the fact that all communication and coordination occurs implicitly through social signaling (chiefly gross body motion, eye gaze, and body language). Furthermore, much of the information people glean about one another’s intentions is derived from the social context. For example, office workers are more likely to be heading towards the cafeteria if it is lunchtime and towards the exit if it is time to go home.
In this talk, I explore some of the mathematical tools that allow us to tease apart the problem of social navigation into patterns that distill enough of the complexity to be learnable. One of the key problems is to predict the future motions of others based on an observed “path prefix”. Past results have shown that geometric prediction of pedestrian motion is nearly impossible to do accurately due to the very fact that people are behaving in a socially competent manner, since they react to other people in ways that achieve their joint goals. Instead, I show how trajectories of navigating pedestrians can be jointly predicted topologically. This prediction can readily be learned in order to understand how people intend to avoid colliding with one another while achieving their goals.