Verifying the Safety of Autonomous Systems with Neural Network Controllers

Simple item page
dc.contributor.authorIvanov, Radoslav
dc.contributor.authorCarpenter, Taylor J.
dc.contributor.authorWeimer, James
dc.contributor.authorAlur, Rajeev
dc.contributor.authorPappas, George
dc.contributor.authorLee, Insup
dc.contributor.authorIvanov, Radoslav
dc.contributor.authorCarpenter, Taylor J.
dc.contributor.authorWeimer, James
dc.contributor.authorAlur, Rajeev
dc.contributor.authorPappas, George
dc.contributor.authorLee, Insup
dc.date2023-05-18T00:18:16.000
dc.date.accessioned2023-05-22T12:51:48Z
dc.date.available2023-05-22T12:51:48Z
dc.date.issued2020-12-01
dc.date.submitted2021-01-04T07:50:54-08:00
dc.description.abstractThis paper addresses the problem of verifying the safety of autonomous systems with neural network (NN) controllers. We focus on NNs with sigmoid/tanh activations and use the fact that the sigmoid/tanh is the solution to a quadratic differential equation. This allows us to convert the NN into an equivalent hybrid system and cast the problem as a hybrid system verification problem, which can be solved by existing tools. Furthermore, we improve the scalability of the proposed method by approximating the sigmoid with a Taylor series with worst-case error bounds. Finally, we provide an evaluation over four benchmarks, including comparisons with alternative approaches based on mixed integer linear programming as well as on star sets.
dc.identifier.urihttps://repository.upenn.edu/handle/20.500.14332/6944
dc.legacy.articleid1911
dc.legacy.fulltexturlhttps://repository.upenn.edu/cgi/viewcontent.cgi?article=1911&context=cis_papers&unstamped=1
dc.source.beginpage1
dc.source.endpage26
dc.source.issue862
dc.source.issue1
dc.source.journalDepartmental Papers (CIS)
dc.source.journaltitleACM Transactions on Embedded Computing Systems (TECS)
dc.source.peerreviewedtrue
dc.source.statuspublished
dc.source.volume1
dc.subject.otherCPS Formal Methods
dc.subject.otherCPS Safe Autonomy
dc.subject.othersoftware and its engineering
dc.subject.otherformal methods
dc.subject.othercomputer systems organization
dc.subject.otherembedded and cyber-physical systems
dc.subject.othercomputing methodologies
dc.subject.otherneural networks
dc.subject.otherComputer Engineering
dc.subject.otherComputer Sciences
dc.titleVerifying the Safety of Autonomous Systems with Neural Network Controllers
dc.typeArticle
digcom.contributor.authorisAuthorOfPublication|email:rivanov@cis.upenn.edu|institution:University of Pennsylvania|Ivanov, Radoslav
digcom.contributor.authorisAuthorOfPublication|email:carptj@seas.upenn.edu|institution:University of Pennsylvania|Carpenter, Taylor J.
digcom.contributor.authorisAuthorOfPublication|email:weimerj@cis.upenn.edu|institution:University of Pennsylvania|Weimer, James
digcom.contributor.authorisAuthorOfPublication|email:alur@cis.upenn.edu|institution:University of Pennsylvania|Alur, Rajeev
digcom.contributor.authorisAuthorOfPublication|email:pappasg@cis.upenn.edu|institution:University of Pennsylvania|Pappas, George
digcom.contributor.authorisAuthorOfPublication|email:lee@cis.upenn.edu|institution:University of Pennsylvania|Lee, Insup
digcom.identifiercis_papers/862
digcom.identifier.contextkey20926732
digcom.identifier.submissionpathcis_papers/862
digcom.typearticle
dspace.entity.typePublication
relation.isAuthorOfPublicationf7d2d91b-2a20-4596-9649-43b222476db4
relation.isAuthorOfPublication520688a9-55b7-4246-9796-704d29020d99
relation.isAuthorOfPublication03b71a10-c612-4bb1-8a80-1463efee44c5
relation.isAuthorOfPublication0277bab3-1be2-4f99-9964-92e0ea36014e
relation.isAuthorOfPublication295f047c-e1ec-4b38-b5f4-0f30e7a34aea
relation.isAuthorOfPublication45a9eed5-3211-4c36-b40d-6394302dfdce
relation.isAuthorOfPublication.latestForDiscoveryf7d2d91b-2a20-4596-9649-43b222476db4
upenn.schoolDepartmentCenterDepartmental Papers (CIS)
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Rado_Postprint.pdf
Size:
3.07 MB
Format:
Adobe Portable Document Format
Download
Collection
Articles