Sequential Decision Making for Intelligent Agents
AAAI Fall Symposium
November 12–14, 2015
Westin Arlington Gateway in Arlington, Virginia adjacent to Washington, DC
Overview of the symposium
Sequential decision making under uncertainty has gained significant traction at major AI conferences such as AAAI, IJCAI, AAMAS (Best Paper award 2014 on this topic) and ICAPS. In many applications, dealing explicitly with uncertainty regarding the effects of actions, state of the environment and possibly the behavior of other agents is crucial to achieve satisfactory task performance. Decision-theoretic planning models like the Markov Decision Process (MDP), the Partially Observable MDP (POMDP) and their many multiagent extensions have emerged as the dominant paradigm for this purpose.
Since its emergence as a subfield of Operations Research, sequential decision making under uncertainty has itself split into many subareas. Within AI, for example, computational aspects of this problem have been analyzed by modelling it under a range of assumptions — as fully observable MDPs, partially observable MDPs, multiagent sequential decision making (MSDM), under full and partial knowledge of parameters, etc. — with a distinct community of researchers focusing on each model. Some of these communities meet each other regularly, but there is no forum dedicated to sequential decision making under uncertainty in AI where all these communities meet simultaneously. This AAAI Fall Symposium aims to bring the researchers of computational sequential decision making under uncertainty together, in the hope that their sharing of ideas with each other will lead to breakthroughs that would be unlikely within any single community on its own.
The format of the symposium is designed to maximize the interaction in a relatively large group of experts in the field. We envision the following elements, which will be grouped together in thematic sessions:
- Invited talks by highly regarded members of one of the sub-communities.
- Shorter contributed talks (see Submission Instructions).
- Panel discussions on specific topics, with a focus on the future of the field.
- Demonstrations of software and tools.
- Plenty of time for breaks to encourage informal discussions.
Invited speakers (titles and abstracts)
- Craig Boutilier, Google
- Emma Brunskill, CMU
- Alan Fern, Oregon State
- Mykel Kochenderfer, Stanford
- Milind Tambe, USC
- Jason Williams, Microsoft Research
- Shlomo Zilberstein, UMass Amherst
Matthijs Spaan (chair), Delft University of Technology, firstname.lastname@example.org
Frans Oliehoek, University of Amsterdam / University of Liverpool, email@example.com
Christopher Amato, University of New Hampshire, firstname.lastname@example.org
Andrey Kolobov, Microsoft Research, email@example.com
Pascal Poupart, University of Waterloo, firstname.lastname@example.org