Option-Aware Adversarial Inverse Reinforcement Learning for Robotic Control

dc.conference.date29 May 2023 - 02 June 2023
dc.conference.locationLondon, United Kingdom
dc.conference.name2023 IEEE International Conference on Robotics and Automation (ICRA)
dc.contributor.authorChen, Jiayu
dc.contributor.authorLan, Tian
dc.contributor.authorAggarwal, Vaneet
dc.contributor.departmentKAUST,CS Department,Thuwal,Saudi Arabia
dc.contributor.departmentComputer, Electrical and Mathematical Science and Engineering (CEMSE) Division
dc.contributor.institutionSchool of Industrial Engineering, Purdue University,West Lafayette,IN,USA,47907
dc.contributor.institutionGeorge Washington University,Department of Electrical and Computer Engineering,Washington D.C.,USA,20052
dc.date.accessioned2023-07-06T07:17:24Z
dc.date.available2022-10-10T11:18:48Z
dc.date.available2023-07-06T07:17:24Z
dc.date.issued2023-07-04
dc.description.abstractHierarchical Imitation Learning (HIL) has been proposed to recover highly-complex behaviors in long-horizon tasks from expert demonstrations by modeling the task hierarchy with the option framework. Existing methods either overlook the causal relationship between the subtask and its corresponding policy or cannot learn the policy in an end-to-end fashion, which leads to suboptimality. In this work, we develop a novel HIL algorithm based on Adversarial Inverse Reinforcement Learning and adapt it with the Expectation-Maximization algorithm in order to directly recover a hierarchical policy from the unannotated demonstrations. Further, we introduce a directed information term to the objective function to enhance the causality and propose a Variational Autoencoder framework for learning with our objectives in an end-to-end fashion. Theoretical justifications and evaluations on challenging robotic control tasks are provided to show the superiority of our algorithm.
dc.eprint.versionPost-print
dc.identifier.arxivid2210.01969
dc.identifier.citationChen, J., Lan, T., & Aggarwal, V. (2023). Option-Aware Adversarial Inverse Reinforcement Learning for Robotic Control. 2023 IEEE International Conference on Robotics and Automation (ICRA). https://doi.org/10.1109/icra48891.2023.10160374
dc.identifier.doi10.1109/icra48891.2023.10160374
dc.identifier.urihttp://hdl.handle.net/10754/682325
dc.publisherIEEE
dc.relation.issupplementedbygithub:LucasCJYSDL/HierAIRL
dc.relation.urlhttps://ieeexplore.ieee.org/document/10160374/
dc.rightsThis is an accepted manuscript version of a paper before final publisher editing and formatting. Archived with thanks to IEEE.
dc.titleOption-Aware Adversarial Inverse Reinforcement Learning for Robotic Control
dc.typeConference Paper
display.details.left<span><h5>Type</h5>Conference Paper<br><br><h5>Authors</h5><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.author=Chen, Jiayu,equals">Chen, Jiayu</a><br><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.author=Lan, Tian,equals">Lan, Tian</a><br><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.author=Aggarwal, Vaneet,equals">Aggarwal, Vaneet</a><br><br><h5>KAUST Department</h5><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.department=KAUST,CS Department,Thuwal,Saudi Arabia,equals">KAUST,CS Department,Thuwal,Saudi Arabia</a><br><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.department=Computer, Electrical and Mathematical Science and Engineering (CEMSE) Division,equals">Computer, Electrical and Mathematical Science and Engineering (CEMSE) Division</a><br><br><h5>Date</h5>2023-07-04</span>
display.details.right<span><h5>Abstract</h5>Hierarchical Imitation Learning (HIL) has been proposed to recover highly-complex behaviors in long-horizon tasks from expert demonstrations by modeling the task hierarchy with the option framework. Existing methods either overlook the causal relationship between the subtask and its corresponding policy or cannot learn the policy in an end-to-end fashion, which leads to suboptimality. In this work, we develop a novel HIL algorithm based on Adversarial Inverse Reinforcement Learning and adapt it with the Expectation-Maximization algorithm in order to directly recover a hierarchical policy from the unannotated demonstrations. Further, we introduce a directed information term to the objective function to enhance the causality and propose a Variational Autoencoder framework for learning with our objectives in an end-to-end fashion. Theoretical justifications and evaluations on challenging robotic control tasks are provided to show the superiority of our algorithm.<br><br><h5>Citation</h5>Chen, J., Lan, T., & Aggarwal, V. (2023). Option-Aware Adversarial Inverse Reinforcement Learning for Robotic Control. 2023 IEEE International Conference on Robotics and Automation (ICRA). https://doi.org/10.1109/icra48891.2023.10160374<br><br><h5>Publisher</h5><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.publisher=IEEE,equals">IEEE</a><br><br><h5>Conference/Event Name</h5><a href="https://repository.kaust.edu.sa/search?spc.sf=dc.date.issued&spc.sd=DESC&f.conference=2023 IEEE International Conference on Robotics and Automation (ICRA),equals">2023 IEEE International Conference on Robotics and Automation (ICRA)</a><br><br><h5>DOI</h5><a href="https://doi.org/10.1109/icra48891.2023.10160374">10.1109/icra48891.2023.10160374</a><br><br><h5>arXiv</h5><a href="https://arxiv.org/abs/2210.01969">2210.01969</a><br><br><h5>Additional Links</h5>https://ieeexplore.ieee.org/document/10160374/<br><br><h5>Relations</h5><b>Is Supplemented By:</b><br/> <ul><li><i>[Software]</i> <br/> Title: LucasCJYSDL/HierAIRL: Codes of the paper: Hierarchical Adversarial Inverse Reinforcement Learning.. Publication Date: 2022-09-13. github: <a href="https://github.com/LucasCJYSDL/HierAIRL" >LucasCJYSDL/HierAIRL</a> Handle: <a href="http://hdl.handle.net/10754/686459" >10754/686459</a></a></li></ul></span>
display.relations<b>Is Supplemented By:</b><br/> <ul><li><i>[Software]</i> <br/> Title: LucasCJYSDL/HierAIRL: Codes of the paper: Hierarchical Adversarial Inverse Reinforcement Learning.. Publication Date: 2022-09-13. github: <a href="https://github.com/LucasCJYSDL/HierAIRL" >LucasCJYSDL/HierAIRL</a> Handle: <a href="http://hdl.handle.net/10754/686459" >10754/686459</a></a></li></ul>
kaust.personAggarwal, Vaneet
orcid.authorChen, Jiayu
orcid.authorLan, Tian
orcid.authorAggarwal, Vaneet
refterms.dateFOA2022-10-10T11:19:27Z
Files
Original bundle
Now showing 1 - 1 of 1
Name:
2210.01969 (1).pdf
Size:
1.03 MB
Format:
Adobe Portable Document Format
Description:
Accepted Manuscript
License bundle
Now showing 1 - 1 of 1
Name:
license.txt
Size:
1.57 KB
Format:
Item-specific license agreed upon to submission
Description:

Version History

Now showing 1 - 2 of 2
VersionDateSummary
2*
2023-07-06 07:16:41
Published as conference paper
2022-10-10 11:18:48
* Selected version