A New Approach to Visual Servoing in Uncalibrated Environments

The objective of the work is to regulate selected features to desired positions on the image plane by controlling motion of a robot manipulator. We assume the system is totally uncalibrated, i.e. both the camera intrinsic parameters and the homogeneous transformation matrix between the robot frame and the vision frame are not calibrated. An adaptive algorithm is proposed to estimate the parameters online. The proposed controller adopts the simple PD plus the gravity compensation scheme with the estimated parameters. A new Lyapunov function is introduced to prove asymptotic convergence of the position errors on the image plane and convergence of the estimated parameters to the real ones up to a scale. Simulations and experiments have been carried out to verify asymptotic stability of the proposed controller