Touché: Data-Driven Interactive Sword Fighting in Virtual Reality

要旨

VR games offer new freedom for players to interact naturally using motion. This makes it harder to design games that react to player motions convincingly. We present a framework for VR sword fighting experiences against a virtual character that simplifies the necessary technical work to achieve a convincing simulation. The framework facilitates VR design by abstracting from difficult details on the lower "physical" level of interaction, using data-driven models to automate both the identification of user actions and the synthesis of character animations. Designers are able to specify the character's behaviour on a higher "semantic" level using parameterised building blocks, which allow for control over the experience while minimising manual development work. We conducted a technical evaluation, a questionnaire study and an interactive user study. Our results suggest that the framework produces more realistic and engaging interactions than simple hand-crafted interaction logic, while supporting a controllable and understandable behaviour design.

キーワード
virtual reality
sword fighting
machine learning
animation
gesture recognition
著者
Javier Dehesa
University of Bath, Bath, United Kingdom
Andrew Vidler
Ninja Theory Ltd., Cambridge, United Kingdom
Christof Lutteroth
University of Bath, Bath, United Kingdom
Julian Padget
University of Bath, Bath, United Kingdom
DOI

10.1145/3313831.3376714

論文URL

https://doi.org/10.1145/3313831.3376714

動画

会議: CHI 2020

The ACM CHI Conference on Human Factors in Computing Systems (https://chi2020.acm.org/)

セッション: Interactions in AR/VR

Paper session
312 NI'IHAU
5 件の発表
2020-04-30 20:00:00
2020-04-30 21:15:00
日本語まとめ
読み込み中…