Towards Data-Driven Sword Fighting Experiences in VR

Abstract

We present a data-driven animated character capable of blocking attacks from a user in a VR sword fighting experience. The system uses motion capture data and a machine learning model to recreate a believable blocking behaviour, suggesting the viability of full-featured data-driven interactive characters in VR. Our work is part of a larger vision of VR interaction as a two-level problem, separating spatial details from design concerns. In this context, here we provide the designers of the experience with a character from which a "blocking'' behaviour can be requested without further spatial specifications. This puts down a first building block in the construction of a controllable data-driven VR sword fighter capable of multiple behaviours.