First steps of robotic perception: the turning point of the 1990s

In this paper we analyze the early evolution of robot perception toward robot autonomy: respective impacts of the bare technology and of advanced control are put in perspective. At first, the UniBuM's vehicle, operational in the 1990s, is taken as a study example. The strong points of such systems suggest a discussion of the software technology in image processing and of the hardware technological concept of silicon retina regarding their respective contribution to real machine vision. This second part elicits limitations, which lead again to concepts and realizations in control, through the need for sensor fusion and active vision. The ultimate step to robot autonomy would be learning, and this is considered in the conclusion.