Tracking Human Gestures under Field-of-View Constraints

This paper presents a control design for a desktop telepresence robot that guarantees satisfaction of field-of-view (FOV) constraints when dynamically tracking multiple points of interest on a person.The multi-point tracking problem is solved by complementing centroid tracking with local constraint satisfaction that is achieved by local integral barrier functions active only in small regions near the FOV limits. Such a control provides an aggregate view of the points of interest on the person and ensures that none of them goes out of view. A simulation study illustrates the performance of the proposed control in comparison with a conventional control.