Abstract

Virtual reality (VR) sickness seems one of the main limitations to the large-scale adoption of VR technologies. This disturbance seems to induce physiological changes that affect the sympathetic and parasympathetic activities of the users. Thereby, it seems relevant to measure users’ physiological data in order to prevent and reduce VR sickness.

This paper presents the results of an initial real-life experiment of VR sickness detection based on physiological data. The electrodermal, cardiac and subjective data of 27 participants was recorded during VR sessions.

Machine learning algorithms were trained and the best model (Gradient Boosting) explained 48% of the VR sickness variance.

These results demonstrate the opportunity to develop an automatic and continuous tool to detect the appearance of VR sickness based on physiological signals. This tool will prove very valuable to the VR industry.

Introduction

Virtual reality (VR) appears as a major technological breakthrough and a main business opportunity for the entertainment industry. The VR market is expected to expand exponentially with worldwide revenues for the AR/VR growing to more than $162 billion in 2020.

However, one main limitation to its large-scale adoption is VR sickness especially because of health, ethical, legal and acceptability aspects.

VR sickness is a common problem that could affect up to 60% of adult users. The necessity to better detect and prevent the appearance of VR sickness is at the origin of this research cooperation between b<>com and Editorial user research lab, Ubisoft Paris aiming at the development of an automatic VR sickness detection tool.