Self-calibration of deformable arm with a monocular camera

A novel robot arm called deformable arm for the service robot is presented in this paper. A flexible working space and the ability to deal with multiple tasks are obtained with a single arm by changing the arm shape. However more flexible use brings difficulties in actual use, the arm's parameters will be changed with the different arm structure and cannot be calibrated with conventional methods. In order to calibrate unknown arm parameters in a short time, a self-calibration method based on known markers attached to the arm and a measurement camera was presented. A rule of attaching markers and a method to get initial parameters of the deformable arm with the changed Denavit-Hartenberg (D-H) model will be discussed first. Then two adjacent calibration poses and positions of the arm were used to establish the relative pose errors model and the cost function. We use the extended Kalman filter (EKF) to optimize kinematic parameters with initial values. Finally, the effectiveness of the proposed method is verified with a deformable arm by the computer simulation and the actual experiment.