Predicting operator performance in smart city applications: an elastic net and ensemble model approach using physiological data from VR simulators

Abstract

This study presents a machine learning framework to predict operator performance in critical smart city applications, such as Urban Air Mobility and autonomous drone control. Using multimodal physiological signals from a public PhysioNet VR flight simulator dataset, we address the high dimensionality of physiological data through advanced feature reduction techniques, including elastic net regularization. We evaluated several ensemble models, with XGBoost combined with elastic net achieving the highest predictive accuracy (MAE=892.95, RMSE=1244.52, R²=0.60). This research demonstrates the potential of physiological monitoring and machine learning for real-time performance assessment of operators in immersive, high-stakes smart city environments, enhancing safety and efficiency in future urban transportation and logistics.