FAPR'96 Workshop

Classical approaches to reasoning about actions and planning are merely concerned with problems involving one central planning agent acting in a strictly static environment, which only changes when the agent itself performs an action. These approaches mostly assume complete and deterministic information about the world state and the effects of actions. Facing more complex domains and realistic applications, however, these assumptions are often inappropriate. For instance, planning problems in robotics are characterized by complex interactions between autonomous units, and uncertainty exists at both planning and execution time. Another crucial aspect of real world problems is that often the concrete outcome of an action or event is not predictable but their relative likelihood can be quantified.

The workshop is intended to respond to this challenge by discussing ontological extensions of existing theoretical formalisms as well as practical solutions in specific application domains. Papers are welcome on all aspects related to this topic including, but not limited to:

• reasoning about concurrent, nondeterministic and continuous actions
• reasoning about actions using probabilistic information, e.g. belief nets
• qualification problem and ramification problem
• planning under uncertainty and with inconsistent information
• distributed AI and planning
• interaction between cooperating autonomous planning agents
• real-time planning
• reusing plans and plan-modification
• real world planning domains and systems, e.g. planning in robotics