Adapting the Point of View for Behavior-Based Navigation

Perception triggered response behavior together with the selection of the appropriate environmental cues can properly guide the robot towards its goal. But often landmarks are difficult to be recognised by the robot; these difficulties are increased due to the morphology of the robot and its movements. Dynamic orientation of the camera’s parameters, fixing the attention of the visual system on relevant environmental features, can greatly improve vision based landmark identification. This paper presents an approach to sonar and compass based dynamic selection of pan, tilt and zoom values.