Challenges and Opportunities in Multiagent Learning for the Real World

Date: 
Monday, 21 March, 2016
Place: 
Palo Alto, CA
Hosting conference: 
AAAI Spring Symposiums
Organizers: 
Christopher Amato University of New Hampshire, USA
Miao Liu Massachusetts Institute of Technology, USA
Karl Tuyls University of Liverpool, UK
Frans Oliehoek University of Amsterdam, NL / University of Liverpool, UK
Jonathan P. How Massachusetts Institute of Technology, USA
Peter Stone University of Texas at Austin, USA
Abstract: 
Developing efficient methods for multiagent learning has been a long-standing research focus in the Artificial intelligence, Game theory, Control, and Neuroscience communities. As a growing number of agents are deployed in complex environments for scientific research and human well-being, there are increasing demands to design efficient learning algorithms that can be used in these real-world settings (including accounting for uncertainty, partial observability, sequential settings and communication restrictions). These challenges exist in many domains, such as underwater exploration, planetary navigation, robot soccer, stock-trading systems, and e-commerce. There has been a lot of great work on multi-agent learning in the past decade, but significant challenges remain, including the difficulty of learning an optimal model/policy from a partial signal, the exploration vs. exploitation dilemma, the scalability and effectiveness of learning algorithms, and convergence guarantees. There are many ways to address these challenges. We are interested in various forms of multiagent learning for this workshop, including: * Learning in sequential settings in dynamic environments (such as stochastic games, decentralized POMDPs and their variants) * Learning with partial observability * Learning with various communication limitations * Learning in ad-hoc teamwork scenarios * Scalability through swarms vs. intelligent agents * Bayesian nonparametric methods for multiagent learning * Deep learning methods for multiagent learning * Transfer learning in multiagent settings * Applications of multiagent learning to real-world systems The purpose of this workshop is to bring together researchers from machine learning, control, neuroscience, robotics, and multi-agent learning/planning communities to discuss how to leverage scalable modern machine learning techniques with the goal of broadening the scope of multi-agent learning research and addressing the fundamental issues that hinder the applicability of multi-agent learning for solving complex real world problems. This workshop will present a mix of invited sessions, contributed talks, poster session by leading experts and active researchers in multi-agent systems, robotics, Bayesian statistics, and decision making and planning. Furthermore, the workshop is designed to allow plenty of time for discussions and initiating collaborations.
Deadline for submission: 
Monday, 19 October, 2015