Advances in Robot Vision: Mechanisms and Algorithms

This paper presents recent results from the Robotics Systems Laboratory at the Australian National University in the area of Robot Vision. Progress has been made in the design and construction of novel mechanisms for both active and panoramic vision, together with computer systems that facilitate real-time vision processing. The robot vision systems are being used to develop human-friendly robots, for guidance in mobile robot navigation and human-machine interaction. 1. ACTIVE VISION Robot vision can be pursued from two contexts; active and passive systems. The biological world has numerous examples of both classes of visual sensors. Active systems have shown that by centering the fovea on the area of interest vision processing tasks such as visual attention, range estimation and tracking can be significantly simplified (Blake & Yuille 1992). A popular approach to building mechanisms for active vision systems has been to provide motor actuation to every degree of freedom. Direct drive systems are fast and have the advantage of zero backlash due to elimination of the gear-train. We exploring a design philosophy based on two principles: ( I ) parallel actuation is preferable to serial actuation because it has the potential to reduce the driven inertia thus reducing motor sizes. Figure 1 shows a typical serial mechanism, where some motors must carry other motors. (2) cable-drive transmission technology is attractive because it provides a stiff, backlash-free reduction that is relatively low-tech and therefore economical. In this paper we describe a prototype pan-tilt mechanism that utilises both principles. The schematic of our prototype active vision system is shown in Figure 2. The independent actuation paths are realised using cable drive technology. The usefulness of cable drive mechanisms for robotics has been proven by (Townsend and Salisbury 1993) in the development of the MIT-WAM robot manipulator. We have used the same cable drive principles to develop our active vision system. A full description of the mechanism is given in (Truong et.al. 1999)