Intent-based Gaze Prediction for Telepresence Robots

The teleoperation of (humanoid) robots allows people to be remotely present to fulfill social or practical tasks. Transmission delays decrease the transparency and usability of such systems, and it is often impossible to decrease these delays to acceptable levels through technological means. By predicting gaze patterns of the user, a robotic avatar can point its camera to where the user will look after one delay’s worth of time, thereby negating the delay. This preliminary work presents a study setup to test such a system.