Abstract

In this paper, we introduce a time-dependent probabilistic map able to model and predict future flow patterns of people in indoor environments. The proposed representation models the likelihood of motion direction by a set of harmonic functions, which efficiently capture long-term (hours to months) variations of crowd movements over time, so from a robotics perspective, this model could be useful to add the predicted human behaviour into the control loop to influence the actions of the robot. Our approach is evaluated with data collected from a real environment and initial qualitative results are presented.