Representational Issues for Real-world Planning Systems

Introduction

In recent years, the AI planning community has begun the transition from artificial problems to realistic applications in space, industrial, and military domains. One lesson from this transition is that successful real-world planning systems require much more than fully automated plan generation.

Plans are created to be executed, often in highly dynamic environments; as such, it is essential for plans to be readily adaptable, possibly in the face of strict deadlines. Interactive and mixed-initiative modalities are preferred by many user communities. The availability of tools to support development of planning knowledge bases and their ongoing maintenance is critical for user acceptance. Users generally want more than one plan that meets their requirements, preferring to have assistance in generating and managing such options. Integration with other systems and knowledge bases is essential, as is the ability for users to comprehend the content and rationale underlying plans.

This workshop brought together researchers and practitioners interested in representational issues for this broader model of planning systems. While there is a rich body of research on core representations for actions and causality, little attention has been paid to the representational requirements needed to support planning-related activities.