Machine learning methods for the study of cybersickness: a systematic review

Area of Science:

• Virtual Reality (VR) and Human-Computer Interaction
• Machine Learning and Artificial Intelligence
• Neuroscience and Biomedical Engineering

Background:

• Virtual Reality (VR) is increasingly integrated into training, therapy, and entertainment, but its use is limited by cybersickness.
• Cybersickness, characterized by nausea, dizziness, and fatigue, is a significant barrier to widespread VR adoption.
• Machine learning (ML) offers a potential solution for identifying and mitigating cybersickness through data analysis.

Purpose of the Study:

• To critically analyze existing machine learning methods for detecting VR-induced cybersickness.
• To review current systems and identify future research directions in ML-based cybersickness analysis.
• To explore the use of biometric and neuro-physiological signals from wearable devices for automated cybersickness identification.

Main Methods:

• Systematic review of 26 selected studies focusing on ML for cybersickness detection.
• Analysis of machine learning methods, data processing techniques, and algorithm architectures.
• Examination of signal acquisition from wearable devices, including biometric and neuro-physiological data.

Main Results:

• Current ML models are effective for detecting cybersickness but lack predictive capabilities for initial occurrences.
• A diverse range of ML architectures, data features, and immersion environments were utilized across studies.
• Biometric and neuro-physiological signals from wearables show promise for automated cybersickness identification.

Conclusions:

• Machine learning holds significant potential for overcoming VR-induced cybersickness limitations.
• Future research should focus on goal-oriented data selection, labeling, and advanced ML models like spiking neural networks.
• Developing predictive models for first-instance cybersickness events is a critical next step.