Influence-Based Abstraction for Multiagent Systems
This paper presents a theoretical advance by which factored POSGs can be decomposed into local models. We formalize the interface between such local models as the influence agents can exert on one another; and we prove that this interface is sufficient for decoupling them. The resulting influence-ba...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | en_US |
| Published: |
Association for the Advancement of Artifical Intelligence
2014
|
| Online Access: | http://hdl.handle.net/1721.1/87052 https://orcid.org/0000-0001-6054-7145 |
| _version_ | 1811094924444041216 |
|---|---|
| author | Oliehoek, Frans A. Witwicki, Stefan J. Kaelbling, Leslie P. |
| author2 | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science |
| author_facet | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Oliehoek, Frans A. Witwicki, Stefan J. Kaelbling, Leslie P. |
| author_sort | Oliehoek, Frans A. |
| collection | MIT |
| description | This paper presents a theoretical advance by which factored POSGs can be decomposed into local models. We formalize the interface between such local models as the influence agents can exert on one another; and we prove that this interface is sufficient for decoupling them. The resulting influence-based abstraction substantially generalizes previous work on exploiting weakly-coupled agent interaction structures. Therein lie several important contributions. First, our general formulation sheds new light on the theoretical relationships among previous approaches, and promotes future empirical comparisons that could come by extending them beyond the more specific problem contexts for which they were developed. More importantly, the influence-based approaches that we generalize have shown promising improvements in the scalability of planning for more restrictive models. Thus, our theoretical result here serves as the foundation for practical algorithms that we anticipate will bring similar improvements to more general planning contexts, and also into other domains such as approximate planning, decision-making in adversarial domains, and online learning. |
| first_indexed | 2024-09-23T16:07:41Z |
| format | Article |
| id | mit-1721.1/87052 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T16:07:41Z |
| publishDate | 2014 |
| publisher | Association for the Advancement of Artifical Intelligence |
| record_format | dspace |
| spelling | mit-1721.1/870522022-10-02T06:29:53Z Influence-Based Abstraction for Multiagent Systems Oliehoek, Frans A. Witwicki, Stefan J. Kaelbling, Leslie P. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Oliehoek, Frans A. Kaelbling, Leslie P. This paper presents a theoretical advance by which factored POSGs can be decomposed into local models. We formalize the interface between such local models as the influence agents can exert on one another; and we prove that this interface is sufficient for decoupling them. The resulting influence-based abstraction substantially generalizes previous work on exploiting weakly-coupled agent interaction structures. Therein lie several important contributions. First, our general formulation sheds new light on the theoretical relationships among previous approaches, and promotes future empirical comparisons that could come by extending them beyond the more specific problem contexts for which they were developed. More importantly, the influence-based approaches that we generalize have shown promising improvements in the scalability of planning for more restrictive models. Thus, our theoretical result here serves as the foundation for practical algorithms that we anticipate will bring similar improvements to more general planning contexts, and also into other domains such as approximate planning, decision-making in adversarial domains, and online learning. United States. Air Force Office of Scientific Research. Multidisciplinary University Research Initiative (Project FA9550-09-1-0538) 2014-05-19T18:27:09Z 2014-05-19T18:27:09Z 2012-07 Article http://purl.org/eprint/type/ConferencePaper http://hdl.handle.net/1721.1/87052 Oliehoek, Frans A, Stefan J. Witwicki, and Leslie P. Kaelbling. "Influence-Based Abstraction for Multiagent Systems." Proceedings of the 26th AAAI Conference on Artificial Intelligence, July 22-26, 2012, Toronto, Ontario, Canada. https://orcid.org/0000-0001-6054-7145 en_US http://www.aaai.org/Conferences/AAAI/2012/aaai12program.pdf Proceedings of the 26th AAAI Conference on Artificial Intelligence Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf Association for the Advancement of Artifical Intelligence MIT web domain |
| spellingShingle | Oliehoek, Frans A. Witwicki, Stefan J. Kaelbling, Leslie P. Influence-Based Abstraction for Multiagent Systems |
| title | Influence-Based Abstraction for Multiagent Systems |
| title_full | Influence-Based Abstraction for Multiagent Systems |
| title_fullStr | Influence-Based Abstraction for Multiagent Systems |
| title_full_unstemmed | Influence-Based Abstraction for Multiagent Systems |
| title_short | Influence-Based Abstraction for Multiagent Systems |
| title_sort | influence based abstraction for multiagent systems |
| url | http://hdl.handle.net/1721.1/87052 https://orcid.org/0000-0001-6054-7145 |
| work_keys_str_mv | AT oliehoekfransa influencebasedabstractionformultiagentsystems AT witwickistefanj influencebasedabstractionformultiagentsystems AT kaelblinglesliep influencebasedabstractionformultiagentsystems |