DMMS 2013 - Tutorial on *Cooperative* Decision Making in Sequential Multiagent Settings

These tutorials will span the range of multiagent interactions of increasing generality, and study a set of optimal and approximate solution techniques to sequential decision making in these multiagent contexts. Recognising the large volume of works, these tutorials are organised in two independent parallel streams: a) the cooperative decision making stream; and b) the adversarial decision making stream. Each stream is self-contained and will begin with the introduction of basic concepts and notions of decision and game theories, and then culminating with several advanced decision-theoretic models of complex agent interactions.
Acronyms covered: MDP, POMDP, Dec-POMDP, I-POMDP, POSG, ND-POMDP, TI-Dec-MDP, OC-Dec-MDP, TD-POMDP.
Description
Choosing optimally among different lines of actions is a key aspect of autonomy in agents. The process by which an agent arrives at this choice is complex, particularly in environments shared with other agents. These tutorials will focus on how to make optimal and approximately optimal decisions in multiagent settings. The will utilise the well-studied domain of search and human support applications to motivate and provide context for a range of multiagent interactions of increasing generality. The focus of this tutorial will be on decision making in sequential interactions, which are often encountered in the search and rescue applications. The tutorial offers two one-day independent streams of lectures with each stream being self-contained, introducing relevant background literature such as aspects of game theory. Stream-I will be dedicated to the cooperative decision making, while Stream-II will address the adversarial issues.
The tutorial is aimed at graduate students and researchers who want to enter this emerging field or to better understand recent results in this area and their implications on the design of multi-agent systems. Participants should have a basic knowledge of probability theory, and preferably, utility theory.
Outline
STREAM I: Cooperative Multiagent Decision Making
Part I: Frameworks for Multiagent Decision Making under Uncertainty Introduction
Search and Rescue Applications in Disaster Management
Requirements for the multiagent decision model and solution
Overview of the basic framework and simple solution methods for them (MDP, POMDP, Bayesian Games)
Part II: Cooperative Models and Algorithms
Dec-POMDP solution concepts
General solution methods
Exploiting structured problems
Other topics (Communication, Learning)
Application problem domains and software tools
STREAM II: Self-Interested Multiagent Decision Making
Part I: Frameworks for Multiagent Decision Making under Uncertainty Introduction
Search and Rescue Applications in Disaster Management
Requirements for the multiagent decision model and solution
Overview of the basic framework and simple solution methods for them (MDP, POMDP, Bayesian Games)
Part II: Self-interested models of decision making
Dynamic algorithms that support equilibrium
Repeated strategic games of complete information
Repeated Bayesian games
Partially Observable Stochastic Games
Modelling and utilising beliefs of others
Interactive POMDPs (I-POMDPs): framework, exact and approximate solution methods, software environments
Emerging applications of multiagent decision making
TTD-MDPs and multiagent Markov tracking