{"id":184500,"date":"2004-05-27T00:00:00","date_gmt":"2009-10-31T13:49:33","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/msr-research-item\/analysis-of-multiagent-teams-using-distributed-pomdps\/"},"modified":"2016-09-09T09:47:05","modified_gmt":"2016-09-09T16:47:05","slug":"analysis-of-multiagent-teams-using-distributed-pomdps","status":"publish","type":"msr-video","link":"https:\/\/www.microsoft.com\/en-us\/research\/video\/analysis-of-multiagent-teams-using-distributed-pomdps\/","title":{"rendered":"Analysis of Multiagent Teams using Distributed POMDPs"},"content":{"rendered":"<div class=\"asset-content\">\n<p>Many current large-scale multiagent team implementations can be characterized as following the \u201cbelief-desire-intention\u201d (BDI) paradigm, with explicit representation of team plans.  Despite their promise, current BDI team approaches lack tools for quantitative performance analyses  under uncertainty. Distributed partially observable Markov decision processes (POMDPs) are well  suited for such analyses, but finding optimal distributed POMDP policies is highly intractable. The key  contribution of this article is a hybrid BDI-POMDP approach, which exploits the positive interactions of these two approaches. In particular, BDI team plans are exploited to improve POMDP tractability and POMDP  analysis improves BDI team plan performance.<\/p>\n<p>Concretely, the structure within the BDI team plans can be exploited to build a factored distributed POMDP model of the domain. The distributed POMDP model can then be used to optimize key decisions within the team plan like role allocation in the presence of uncertainty. Here again the structure within the team plan can be exploited to improve the tractability of the optimization step. Further, the belief-based decision making in BDI plans yields a more efficient representation of policies and a significantly  faster policy evaluation algorithm suited for our BDI-POMDP hybrid approach. As my research highlights, using a hybrid BDI-POMDP approach allows for analyzing multiagent teams in the presence of uncertainty. Exploiting the positive interactions between the two approaches helps reduce the native intractability of distributed POMDPs, enabling their use in complex dynamic domains. I will demonstrate the benefit of using this hybrid approach in two key domains. One of the domains is  RoboCupRescue, where I will illustrate the significant practical improvements in  allocating teams in disaster rescue simulations.<\/p>\n<\/div>\n<p><!-- .asset-content --><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Many current large-scale multiagent team implementations can be characterized as following the \u201cbelief-desire-intention\u201d (BDI) paradigm, with explicit representation of team plans. Despite their promise, current BDI team approaches lack tools for quantitative performance analyses under uncertainty. Distributed partially observable Markov decision processes (POMDPs) are well suited for such analyses, but finding optimal distributed POMDP policies [&hellip;]<\/p>\n","protected":false},"featured_media":195457,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"footnotes":""},"research-area":[],"msr-video-type":[],"msr-locale":[268875],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-184500","msr-video","type-msr-video","status-publish","has-post-thumbnail","hentry","msr-locale-en_us"],"msr_download_urls":"","msr_external_url":"https:\/\/youtu.be\/CX-fA1K3i5M","msr_secondary_video_url":"","msr_video_file":"","_links":{"self":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/184500"}],"collection":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video"}],"about":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-video"}],"version-history":[{"count":0,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video\/184500\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media\/195457"}],"wp:attachment":[{"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/media?parent=184500"}],"wp:term":[{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=184500"},{"taxonomy":"msr-video-type","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-video-type?post=184500"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=184500"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=184500"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/www.microsoft.com\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=184500"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}