A Planning Algorithm for Incorporating Attempts and Nondeterminism into Interactive Stories

In order to maintain their replay value and entertain the users, interactive storytelling systems demand the support of algorithms that, besides providing diversity of plots (without losing coherence) and possibilities of interaction, are capable of generating stories in a suitable amount of time, so that the user experience doesn't result in frustration. In this context, automated planning algorithms are interesting alternatives, as they can create multiple plots that are coherent with the intended genre. A well established approach used to improve the performance of planners is the use of HTN planning techniques. In this paper, we present NDetHPlan, a general purpose planning algorithm that combines HTN and nondeterminism to cope with such requirements. The planner presented here can also deal with failed attempts to achieve goals, an especially valuable feature to create dramatic tension in the context of interactive storytelling. NDetHPlan has been incorporated to Log tell, a logic-based tool for interactive generation and dramatization of stories.