Multiagent planning is concerned with planning by (and for) multiple agents. It can involve agents planning for a common goal, an agent coordinating the plans or planning of others, or agents refining their own plans while negotiating over tasks or resources. Distributed continual planning addresses these problems when further complicated with interleaved execution. More than ever industry, space, and the military are seeking systems that can solve these problems.
This tutorial will describe variations of the multiagent planning problem, discuss issues in the applicability and design of multiagent planning systems, and describe some real-world multiagent planning problems. We will also review the history of research contributions to this sub-field and describe frameworks and systems such as Distributed NOAH, GPGP, DSIPE, and SHAC. In addition, we will describe open research issues in multiagent planning and its overlap and relation to other fields, such as market-based AI and game theory.
Tentative tutorial outline:
Basic knowledge of artificial intelligence and planning techniques will be helpful, but not necessary. This tutorial will give researchers and practitioners an understanding of the motivations, applications, and history of work in multiagent planning up to present day. After this tutorial, a graduate student could choose a thesis topic and know how to situate it with prior work. A research practitioner or systems engineer would have references to relevant research and resources to implement a multiagent planning system.
Bradley J. Clement
Brad Clement is a senior member of the Artificial Intelligence
Group at the Jet Propulsion Laboratory in Pasadena, CA, where he
is developing methods for coordinating planning and scheduling for
single and multiple spacecraft/missions. He leads projects on
distributed continual planning applied to simulated spacecraft and
rovers for Mars, on scheduling resource allocation for the Deep
Space Network, and on planning under uncertainty. He received a
NASA Space Act Award for work on obtaining exponential speedups in
using abstraction in planning and scheduling applied to a rover
team. He received a Ph.D. degree in computer science and
engineering from the University of Michigan, Ann Arbor. His
research interests include multiagent coordination, situated
planning and execution, distributed systems, and AI in games.
Keith S. Decker
Keith Decker is an Associate Professor in the Department of
Computer and Information Sciences at the University of Delaware.
His research interests include cooperative distributed problem
solving, multiagent systems, computational organization design,
real-time AI, parallel and distributed planning and scheduling,
and distributed information gathering. He received a National
Science Foundation CAREER Award in 1998, is a member of the
editorial board for the Autonomous Agents and Multiagent Systems
Journal, and has been publicity and workshop program chair for
several Agent conferences, including the first International
Conference on Autonomous Agents and Multiagent Systems in 2002.
Dr. Decker's contributions to artificial intelligence center
around coordination in multiagent systems. TAEMS was developed by
Decker to represent complex coordination problems, and GPGP as one
approach to solving them. Keith's group is also responsible for a
complete, fielded agent internal architecture and agent
development toolkit called DECAF, and is currently working on
models of mixed human and computational agent organizations and
using agent-based systems for information gathering in
bioinformatics.