Guiding attention for grasping tasks by gestural instruction: the GRAVIS-robot architecture | Synapse