A VISUAL SERVOING APPROACH FOR ROBOTIC FRUIT HARVESTING IN THE PRESENCE OF PARAMETRIC UNCERTAINTIES
Abstract
In this paper, we address the robust visual servoing problem for a fixed target using an eyein-
hand camera configuration, when the camera calibration parameters are uncertain. The proposed solution
combines the image-based visual servoing (IBVS) and the position-based visual servoing (PBVS) approaches to
take advantage of both strengths to perform successful robotic fruit harvesting tasks. An image-based detection and recognition method, based on SURF and RANSAC algorithms, is used to extract the image feature of the fruit for the vision-based control algorithm. A kinematic control design, with robustness properties, is employed to cope with the uncertainties in the calibration parameters of the camera-robot system. To deal with possible singular configurations of the robot arm, that may arise during the task execution, we employ the Damped Least-Squares (DLS) inverse method. Experimental results, obtained with a Mitsubishi robot arm RV-2AJ carrying out strawberry harvesting tasks, are included to illustrate the performance and feasibility of the proposed methodology.