Performatology: A Procedural Acting Approach for Interactive Drama in Cinematic Games

In this paper we define Performatology as a data driven, machine learning, and performance theory based approach to interactive drama through the procedural acting of Performative Embodied Agents (PEAs). PEAs are designed to serve as acting tools for live embodied performers to extend their craft into computational media, and to make agents better actors for the enhancement of cinematic games. We position Performatology as a complement to Narratology research, as cinematic games require good procedural actors to dramatically tell their stories. The design of our PEAs architecture features a Mimic module that simulates a live actor’s gesture through performer modeling, recognition, prediction, and synthesis. To model performer presence, we initially implement offline training of a genetic algorithm on mocap data to learn a dancer’s gestural signature, and then use the model for online interaction with the same performer using a Kinect camera.