This is an old revision of the document!
May 5-6, 2014 in Paris, France
In sequential decision making, an agent's objective is to choose actions, based on observations of its environment, that will maximize the expected performance over the course of a series of such decisions. In worlds where action consequences are non-deterministic or observations incomplete, Markov Decision Processes (MDPs) and Partially-Observable MDPs (POMDPs) serve as the basis for principled approaches to single-agent sequential decision making. Extending these models to systems of multiple agents has become the subject of an increasingly active area of research. Over the past decade, a variety of different multiagent models have emerged, both for fully-cooperative agents (e.g., MMDP, MTDP and Dec-POMDP) and self-interested agents (e.g., I-POMDP and POSG), and under an assortment of different assumptions about agents' capabilities to communicate (e.g., Dec-MDP-Com, COM-MTDP), observe (e.g., Dec-MDP) and influence other agents (e.g., TI-Dec-MDP, ND-POMDP). Often, high computational complexity has driven researchers to develop multiagent planning and learning methods that exploit structure in agents' interactions, methods geared towards efficient approximate solutions, decentralized methods that distribute computation among the agents, and new ways for agents to model and reason about their interactions with other agents.
The purpose of this workshop is to bring together current and future researchers in the field of multiagent sequential decision making (MSDM) to present and discuss promising new work, to identify recent trends in model and algorithmic development, and to establish important directions and goals for further research and collaboration. This workshop also strives to develop consensus within the community on benchmarks and evaluation methodology in order to compare and validate alternative approaches and models. Further, we hope that these active discussions and collaborations will help us to overcome the challenges of successfully applying MSDM methods to real-world problems in security, sustainability, public safety and health, and other challenging domains.
This year, the MSDM workshop will also include an extensive tutorial geared towards introducing the fundamental concepts of multiagent sequential decision making, acclimating researchers to the broad landscape of MSDM models and methods, and informing potential practitioners of the state of the art.
Multiagent sequential decision making comprises (1) problem representation, (2) planning, (3) coordination, and (4) learning. The MSDM workshop addresses this full range of aspects. Topics of particular interest include: