Automatically staging characters in order to facilitate the performance and blocking of a scene in a virtual environment is a difficult task today. There is only a limited set of techniques used in practice. Some general methods include Behavior Markup Language (BML) and motion capture replay (the most popular); however they require either detailed technical knowledge or are not adaptable to different environmental configurations. In this work, we block and perform scenes with synthetic actors utilizing only a text-based, standard play-script as the primary input to the positioning of these characters in an environment. Using natural language processing techniques, we extract the annotated movements from the script, then add additional movements and adjustments from our rules engine built with theater, stage performance, and human interaction spatiotemporal relationships. In addition, we incorporate force-directed graph algorithms to adjust positions of the artificially intelligent (AI)-driven characters based on human-controlled character movements for interaction. These techniques have been quantitatively and qualitatively evaluated, revealing both similar blocking, and indistinguishably good performances from a human's perspective, when compared to an actual human performance.

中文翻译：

---

利用 Play-Script 时空推理的虚拟环境定位

自动上演角色以促进虚拟环境中场景的表现和阻塞是当今的一项艰巨任务。在实践中只使用了有限的一组技术。一些通用方法包括行为标记语言 (BML) 和动作捕捉回放（最流行的）；然而，它们要么需要详细的技术知识，要么不能适应不同的环境配置。在这项工作中，我们仅使用基于文本的标准剧本作为在环境中定位这些角色的主要输入，与合成演员一起阻止和表演场景。使用自然语言处理技术，我们从脚本中提取带注释的动作，然后从我们的规则引擎中添加额外的动作和调整，该引擎是由剧院、舞台表演、和人际交往的时空关系。此外，我们结合了力导向图算法，根据人类控制的交互动作来调整人工智能 (AI) 驱动的角色的位置。这些技术已经过定量和定性评估，与实际人类表现相比，从人类的角度揭示了类似的阻塞和难以区分的良好表现。