gaips_bea image1 image2 image3 image4 image5 gaips_ecute_beach_bar_banner gaips_ecute_train_incorrect_ticket_banner
Influence-Based Abstraction for Multiagent Systems


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.
Year 2012
Keywords Influence Abstraction, Multiagent Planning and Learning, Sequential Decision Making
Authors Frans Oliehoek, Stefan Witwicki, Leslie Kaelbling
Booktitle Proceedings of the Twenty-Sixth AAAI Conference (AAAI-2012)
Address Toronto, Canada
Month July
Pdf File
BibTex bib icon or see it here down icon

@inproceedings { oliehoek12, 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.}, address = {Toronto, Canada}, booktitle = {Proceedings of the Twenty-Sixth AAAI Conference (AAAI-2012)}, keywords = {Influence Abstraction, Multiagent Planning and Learning, Sequential Decision Making}, month = {July}, title = {Influence-Based Abstraction for Multiagent Systems}, year = {2012}, author = {Frans Oliehoek and Stefan Witwicki and Leslie Kaelbling} }

up icon hide this content