We propose an autonomous vision system architecture, which consists of a reactive control scheme and an active vision scheme, for automatically building maps of unknown environments. The validity of this architecture is demonstrated through a hardware model experiment.
