Call for Papers
We solicit papers relevant to single or multi-agent decision making under
uncertainty and partial observability, including (though not limited to) the following:
- Leveraging techniques and ideas from research on planning, such as employing relaxations,
and replanning, for partially observable domains.
- Extending work on abstractions and generalized solution methods with theoretical
guarantees to partially observable and multiagent domains.
- Transforming ideas and intuitions between the neighboring communities. We welcome
papers that discuss the relationships between the communities and discuss even very early
- While multiagent planning problems are typically much harder to solve, defining or learning
the models for the agent(s) is of comparable level of difficulty. Having good models at an
appropriate level of abstraction is important for real-world applications. We welcome
contributions on how to learn appropriate stochastic models, or on related practical issues.
- Often planning under uncertainty problems arise in hybrid domains whose dynamics include
both a stochastic and a deterministic component, or both partially observable and observable
state variables. Prior work has demonstrated that approaches that leverage the reduced
uncertainty in the resulting problem can gain computational benefits. We also welcome
contributions exploring similar simplifications of multi-agent problems that result in
computational speed ups or improved performance quality.
- Making decisions in multi-agent systems naturally raises the question of whether the agents
are cooperative or competitive, and both areas are active centers of research. We invite work
that considers the possible symmetry between various planning criteria in single-agent
systems and collaborative/competitive scenarios in multi-agent systems, and how such a
symmetry may lead to new research advances in both communities.
Submissions that discuss the relationships between conformant planning, POMDPs and/or DEC-POMDPs
will be given priority in acceptance. Please note that we also welcome papers that are also under review (or recently accepted) for upcoming conferences. Such papers may be submitted as a pdf with the authors' names left off, and will be reviewed by reviewers unaware of the authors' identity, if that is so wished by the authors.