Computer Science > Robotics
[Submitted on 13 Jan 2015 (v1), revised 23 Jan 2015 (this version, v2), latest version 29 Apr 2015 (v3)]
Title:Localizing antipodal grasps in point clouds
View PDFAbstract:This paper proposes a new approach to localizing and grasping novel objects in clutter. The input to our algorithm is a point cloud and the geometric parameters of the robot hand. The algorithm locates hand configurations that are expected to result in good grasps. One of the challenges we address is that real-life point clouds are typically missing lots of data as a result of occlusions and object surface specularities. This can make perception hard. Our key innovation is to frame the grasp perception problem in terms of predicting the presence of antipodal grasps. The advantage of this approach is that an antipodal grasp is an objective and verifiable goal for localization that enables us to automate labeling of the training set. The result is larger training sets, less noisy labels, and better cross-validation that ultimately gives us better grasp prediction performance. Our experiments indicate that our approach will successfully grasp on average 96% of novel objects presented to the robot in a single-object scenario. In experiments where items are grasped from a disordered pile of 26 objects, our approach is able to grasp 90% of the objects successfully.
Submission history
From: Andreas ten Pas [view email][v1] Tue, 13 Jan 2015 18:40:11 UTC (5,529 KB)
[v2] Fri, 23 Jan 2015 20:36:56 UTC (8,370 KB)
[v3] Wed, 29 Apr 2015 02:22:38 UTC (5,551 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.