Influence-Based Abstraction for Multiagent Systems (bibtex)
by Frans A. Oliehoek and Stefan Witwicki and Leslie P. Kaelbling
Abstract:
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.
Reference:
Influence-Based Abstraction for Multiagent Systems (Frans A. Oliehoek and Stefan Witwicki and Leslie P. Kaelbling), In Proceedings of the National Conference on Artificial Intelligence, 2012.
Bibtex Entry:
@InProceedings{Oliehoek12AAAI_IBA,
 author = {Frans A. Oliehoek and 
 Stefan Witwicki and
 Leslie P. Kaelbling},
 title = {Influence-Based Abstraction for Multiagent Systems},
 booktitle = {Proceedings of the National Conference on Artificial Intelligence},
 month = jul,
 year = 2012,
 OPTpages = {},
keywords={Multiagent},
 bib2html_rescat = {Multiagent systems - decentralized (approximate) planning under uncertainty},
 bib2html_pubtype = {Refereed Conference (International)},
 abstract =  {
 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.
 },
url={http://people.csail.mit.edu/fao/docs/Oliehoek12AAAI_IBA.pdf}
}
Powered by bibtexbrowser