Visual servoing from uncalibrated cameras for uncalibrated robots

Visual servoing is very useful for navigating robots to specific positions and orientations accurately and reliably. Unfortunately, the existing methods suffer from the requirement of camera calibrations as well as robot calibrations. In fact, it is very difficult to calibrate cameras and robots accurately. In this paper, we consider epipolar geometry and show that even if the cameras and robots are uncalibrated, visual servoing can be achieved from epipolar geometry. In particular, we show how epipolar geometry can be exploited for computing relative orientations between the direction of translations and the goal positions. We present preliminary results from uncalibrated visual servoing as well as statistical evaluations to show the power of the proposed method.