File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Efficient penetration depth approximation using active learning

TitleEfficient penetration depth approximation using active learning
Authors
KeywordsContact Space
Dynamic Simulation
Penetration Depth
Support Vector Machine
Active Learning
Issue Date2013
Citation
ACM Transactions on Graphics, 2013, v. 32, n. 6 How to Cite?
AbstractWe present a new method for efficiently approximating the global penetration depth between two rigid objects using machine learning techniques. Our approach consists of two phases: offline learning and performing run-time queries. In the learning phase, we precompute an approximation of the contact space of a pair of intersecting objects from a set of samples in the configuration space. We use active and incremental learning algorithms to accelerate the precomputation and improve the accuracy. During the run-time phase, our algorithm performs a nearest-neighbor query based on translational or rotational distance metrics. The run-time query has a small overhead and computes an approximation to global penetration depth in a few milliseconds. We use our algorithm for collision response computations in Box2D or Bullet game physics engines and complex 3D models and observe more than an order of magnitude improvement over prior PD computation techniques.
Persistent Identifierhttp://hdl.handle.net/10722/206224
ISSN
2015 Impact Factor: 4.218
2015 SCImago Journal Rankings: 2.552
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorPan, Jia-
dc.contributor.authorZhang, Xinyu-
dc.contributor.authorManocha, Dinesh-
dc.date.accessioned2014-10-22T01:25:29Z-
dc.date.available2014-10-22T01:25:29Z-
dc.date.issued2013-
dc.identifier.citationACM Transactions on Graphics, 2013, v. 32, n. 6-
dc.identifier.issn0730-0301-
dc.identifier.urihttp://hdl.handle.net/10722/206224-
dc.description.abstractWe present a new method for efficiently approximating the global penetration depth between two rigid objects using machine learning techniques. Our approach consists of two phases: offline learning and performing run-time queries. In the learning phase, we precompute an approximation of the contact space of a pair of intersecting objects from a set of samples in the configuration space. We use active and incremental learning algorithms to accelerate the precomputation and improve the accuracy. During the run-time phase, our algorithm performs a nearest-neighbor query based on translational or rotational distance metrics. The run-time query has a small overhead and computes an approximation to global penetration depth in a few milliseconds. We use our algorithm for collision response computations in Box2D or Bullet game physics engines and complex 3D models and observe more than an order of magnitude improvement over prior PD computation techniques.-
dc.languageeng-
dc.relation.ispartofACM Transactions on Graphics-
dc.subjectContact Space-
dc.subjectDynamic Simulation-
dc.subjectPenetration Depth-
dc.subjectSupport Vector Machine-
dc.subjectActive Learning-
dc.titleEfficient penetration depth approximation using active learning-
dc.typeArticle-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1145/2508363.2508305-
dc.identifier.scopuseid_2-s2.0-84887848946-
dc.identifier.volume32-
dc.identifier.issue6-
dc.identifier.eissn1557-7368-
dc.identifier.isiWOS:000326923200035-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats