Publication Type

Conference Proceeding Article

Publication Date



Multiagent sequential decision making has seen rapidprogress with formal models such as decentralized MDPsand POMDPs. However, scalability to large multiagent systemsand applicability to real world problems remain limited.To address these challenges, we study multiagent planningproblems where the collective behavior of a populationof agents affects the joint-reward and environment dynamics.Our work exploits recent advances in graphical modelsfor modeling and inference with a population of individualssuch as collective graphical models and the notion of fi-nite partial exchangeability in lifted inference. We developa collective decentralized MDP model where policies can becomputed based on counts of agents in different states. Asthe policy search space over counts is combinatorial, we developa sampling based framework that can compute openand closed loop policies. Comparisons with previous best approacheson synthetic instances and a real world taxi datasetmodeling supply-demand matching show that our app


Computer and Systems Architecture | Databases and Information Systems | Systems Architecture

Research Areas

Information Systems and Management


AAAI Conference on Artificial Intelligence (AAAI): San Fransisco, USA, 2017 February 4

First Page


Last Page




City or Country

San Fransisco, USA

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.