Immediate Detection of Simulator Sickness in Virtual Environments Using Integrated Subjective Feedback and Physiological Signals

Simulator sickness poses a significant challenge to the effectiveness of training programs in virtual environments. Early diagnosis of this syndrome can significantly prevent its onset and enhance user safety by enabling the management of symptoms. In our study, thirty-one participants were exposed to a recorded video of a flight in the Helimond simulator. Simulator sickness was then quantified using the Simulator Sickness Questionnaire, a Forced Binary Choice method, and an Equivital sensor. We analyzed the correlations between physiological signals and self-assessment scores, as well as evaluated the performance of various machine learning algorithms in predicting simulator sickness. The results indicated a strong correlation between SSQ scores and forced choice responses. Among the machine learning algorithms tested, Random Forest outperformed the others, achieving the highest accuracy (0.937) and AUC (0.932). Decision Trees followed in performance, while Logistic Regression showed the worst performance. Furthermore, based on our dataset, heart rate and Inter-Beat Interval were identified as the most significant factors for predicting simulator sickness. Our approach emphasizes the importance of instant assessment and prediction of simulator sickness, providing a simple yet effective method to improve the evaluation process. This method facilitates prompt evaluation and proactive symptom prevention, benefiting virtual environment design and user safety.