Active vision gaze control for mobile robot

Active vision systems for use in robots require a method of directing the camera at targets of interest. There are two approaches to solving this problem; a bottom-up approach which directs the camera at elements of the visual field that have high salience, or a top-down approach which directs the vision under deliberative control. We discuss a biologically inspired system that combines bottom-up and top-down approaches. An artificial analogue of the superior colliculus is used to combine a set of data-driven feature maps. The relative weighting of the various maps are dependent on the task being performed by the robot. In this paper we discuss the architecture of the gaze control system and present simulations illustrating the operation of the artificial superior colliculus.