Active vision models can simplify visual tasks, provided that they can select sensible actions given incom-ing sensory inputs. Many active vision models have been proposed, but a comparative evaluation of these models