An actively altruistic mobile robot system for identifying someone who looks lost using head pose tracking

We present an actively altruistic mobile robot system that implements a neural network to perform head pose tracking to identify a person who looks lost. Our results show that a 3-level discretization of head pose performed better than 9 levels, but that 9 levels perform better at detecting head pose change. We also present results from analyzing a method that determines whether or not a person looks lost by using head pose change frequency from the most recent 20 frames.