Planning High-Quality Grasps Using Mean Curvature Object Skeletons

Abstract

In this work, we present a grasp planner which integrates two sources of information to generate robust grasps for a robotic hand. First, the topological information of the object model is incorporated by building the mean curvature skeleton and segmenting the object accordingly in order to identify object regions which are suitable for applying a grasp. Second, the local surface structure is investigated to construct feasible and robust grasping poses by aligning the hand according to the local object shape. We show how this information can be used to derive different grasping strategies, which also allows to distinguish between precision and power grasps. We applied the approach to a wide variety of object models of the KIT and the YCB real-world object model databases and evaluated the approach with several robotic hands. The results show that the skeleton-based grasp planner is capable to autonomously generate high-quality grasps in an efficient manner. In addition, we evaluate how robust the planned grasps are against hand positioning errors as they occur in real-world applications due to perception and actuation inaccuracies. The evaluation shows that the majority of the generated grasps are of high quality since they can be successfully applied even when the hand is not exactly positioned.

Modelling the structure of object-independent human affordances of approaching to grasp for robotic hands

“…Hence, if the environment does play a role in adjusting the grasp pattern, a soft hand could provide a model-free grasp finalisation to complement our model. Additionally, such pairing could drastically reduce the use of ad-hoc precomputed grasp affordances or approaching profiles [74]. Also, the computational complexity of object perception algorithms [75], which often rely on expensive and power intensive ad-hoc processors (i.e.…”

Attention Based Visual Analysis for Fast Grasp Planning With a Multi-Fingered Robotic Hand

“…Previous grasp planning methods can be divided into geometric-based grasping and similarity-based grasping. In geometric-based grasping (Hsiao et al, 2010; Laga et al, 2013; Vahrenkamp et al, 2018), geometric information of the object is obtained from color or depth images, and it is used to define a set of heuristics to guide grasp planning. Hsiao et al (2010) proposed a heuristic which maps partial shape information of objects to grasp configuration.…”

“…In the computer vision community, most previous works sample human hand pose with a motion tracking system and use it to detect hand grasp types (Rogez et al, 2015; Cai et al, 2017). In the robotics community, there are few previous approaches that try to integrate grasp type detection into robotic grasp planning (Harada et al, 2008; Vahrenkamp et al, 2018). In those works, only two kinds of grasp types, i.e., power and precision (Napier, 1956), are considered, which is not sufficient for exploring the potential of multi-fingered robot hands.…”

“…Moreover, the desired grasp type is determined manually for robotic hands. In terms of visual analysis, there are approaches which use visual analysis to define heuristics or constraints for grasp planning (Hsiao et al, 2010; Aleotti and Caselli, 2012; Vahrenkamp et al, 2018). In comparison to those approaches, there are three main differences: (1) our approach learns features directly from raw sensor data, while most of the previous approaches use handcrafted features; (2) six grasp types are considered while the previous approaches only consider two grasp types.…”

optimization Model for Planning Precision Grasps with Multi-Fingered Hands

“…However, the Reeb graph is slow to build, and it assumes only one part will be grasped for on object, whereas the optimal grasp can expand on multiple parts, especially for small objects. In [5], the grasps were sampled from the mean curvature skeleton of the object with the manuallydesigned heuristics. The skeleton is computationally heavy, and the collision was addressed by simply moving backward against the approaching direction.…”