Another look at search-based drama management

作者: Mark J. Nelson , Michael Mateas

DOI:

关键词:

摘要: A drama manager (DM) monitors an interactive experience, such as a computer game, and intervenes to shape the global experience so it satisfies author's expressive goals without decreasing player's agency. In declarative optimization-based management (DODM), author declaratively specifies desired properties of experience; DM optimizes its interventions maximize that metric. The initial DODM approach used online search optimize experience-quality function. Subsequent work questioned whether could perform well in general, proposed alternative optimization frameworks reinforcement learning. Recent on targeted trajectory distribution Markov decision processes (TTD-MDPs) replaced metric with associated algorithm based targeting distributions. We argue optimizing function does not destroy agency, has been claimed, fact can capture goal directly. further show that, though apparently quite different surface, original TTD-MDPs actually use variants same underlying algorithm, offline cached search, is done by TTD-MDP allows search-based systems achieve similar results TTD-MDPs.

参考文章(0)