Robust feature extraction and control design for autonomous grasping and mobile manipulation

This paper presents a novel design of visual servo control of a mobile manipulator for autonomous grasping of a target object. In this design, scale invariant feature transform (SIFT) algorithm is adopted to search and recognize the object to grasp. Random sample consensus (RANSAC) algorithm is used to remove outliers and find the refined homography matrix between database and current image. Robust feature matching provides reliable feature points to the image-based visual servo control loop. Experimental results show that the mobile manipulator can find and grasp a target object autonomously using the proposed method.