Autonomous switching of top-down and bottom-up attention selection for vision guided mobile robots.