Using Experience for Assessing Grasp Reliability

Abstract

Abstract. Manipulation skills are a key issue for a humanoid robot. Here, we are interested in a vision-based grasping system able to deal with previously unknown objects in real time and in an intelligent manner. Starting from a number of feasible candidate grasps, we focus on the problem of predicting their reliability using the knowledge acquired in previous grasping experiences. A set of visual features which take into account physical properties that can affect the stability and reliability of a grasp are defined. A humanoid robot obtains its grasping experience by repeating a large number of grasping actions on different objects. An experimental protocol is established in order to classify grasps according to their reliability. Two prediction/classification strategies are defined which allow the robot to predict the outcome of a grasp only analizing its visual features. The results indicate that these strategies are adequate to predict the realibility of a grasp and to generalize to different objects.

Towards Grasp-Oriented Visual Perception for Humanoid Robots

“…In contrast to that Goodale et al, 59 Cuijpers et al 60 and Gentilucci 61 emphasize the importance of global object shape for the purpose of reaching and grasping in humans. The work by Morales et al 52 applies these findings on the relevance of global object shape to robotic grasping. Here, also the hand kinematics are considered to infer a number of planar grasp configurations directly from 2D object contours obtained through vision.…”

“…Saxena et al, 51 and Morales et al 52 and Stark et al 53 apply monocular images to derive suitable grasps. Such 2D approaches avoid the difficult problem of 3D reconstruction, as also their applicability is supported by a number of articles in the field of neurophysiology.…”

“…The approach is different from the one taken by Saxena et al, 51 Stark et al 53 where only local appearance or affordance cues are used. In contrast to Morales et al 52 where only planar objects are considered, we are considering arbitrarily shaped novel objects.…”

An Active Vision System for Detecting, Fixating and Manipulating Objects in the Real World

“…More recently, it has been proposed to use vision as a solution to obtain the lacking information about object shapes or to use contact information to explore the object Morales et al, 2001;Platt Jr. et al, 2002). Another trend has focused on the use of machine learning approaches to determine the relevant features that indicate a successful grasp (Coelho et al, 2001;Kamon et al, 1998;Morales et al, 2004). Finally, there have been efforts to use human demonstrations for learning grasp tasks (Ekvall and Kragic, 2004).…”

Movement Speed Models of Natural Grasp and Release Used for an Industrial Robot Equipped with a Gripper

“…The work on automatic grasp synthesis and planning is a relevant idea [6,7,8]. In [10] is presented a method for automatic grasp based on shape primitives of an object, using a programming by demonstration platform.…”