Agents' Interaction in Virtual Storytelling

In this paper we describe a fully implemented prototype for interactive storytelling using the UnrealTM engine. Using a sit-com like scenario as an example of how the dynamic interactions between agents and/or the user dramatise the emerging story. Hierarchical Task Networks (HTNs) are formalised using AND/OR graphs, which are used to describe the many possible variations of the story at a sub-goal level, and the set of all behaviours (from a narrative perspective) of the primary actors at a terminal action level. We introduce real-time variant of the heuristic search algorithm AO* that has been implemented to provide a mechanism for planning (and re-planning) and discuss how the chosen heuristic evaluation function is used to describe narrative concepts. We provide early results of several examples of how the same basic plot can have many differing story instantiations as a result of the dynamic interaction within the virtual set and the personalities of the primary characters, and detail the steps required in the plan generation.