Using estimation techniques for visual servoing in dynamic environments

This paper is concerned with the visual servoing of a mobile robot in dynamic environments. We assume a target with maneuvering capabilities, and thus it can be hidden from the camera by the obstacles in the scene. These two problems must be taken into account in the control law to ensure correct servoing. The control law must consider the target movement to reduce tracking error as small as possible. Moreover, the control law should consider visual loss management (reconstruction of the visual signal in case of occultation), and collision avoidance, estimating the obstacle motion. We present a strategy to avoid the tracking error due to the movement of the target itself.