Combined Method Comprising Low Burden Physiological Measurements with Dry Electrodes and Machine Learning for...

Area of Science:

• Construction Engineering
• Human Factors Engineering
• Biomedical Engineering

Background:

• Remote-controlled excavators are crucial for addressing labor shortages and enhancing safety in construction.
• Visually induced motion sickness (VIMS) poses a significant challenge to the effective remote operation of heavy machinery.

Purpose of the Study:

• To develop and validate a prototype system for predicting the occurrence and severity of VIMS in remote excavator operators.
• To identify key physiological indicators correlated with VIMS during simulated heavy machinery operation.

Main Methods:

• Acquired multiple low-burden physiological signals (e.g., saccade frequency, skin conductance) from nine participants operating excavator simulators.
• Constructed and validated Light Gradient-Boosting Machine (LGBM) binary classification models using approximately 30,000 seconds of time-series data and 23 features.
• Evaluated model performance using leave-one-out cross-validation and Area Under the Curve (AUC) metrics.

Main Results:

• Achieved a mean Receiver Operating Characteristic (ROC) AUC score of 0.84 and a mean Precision-Recall (PR) AUC score of 0.71.
• Identified saccade frequency and skin conductance response as particularly important features for VIMS detection.
• Model performance correlated well with subjective VIMS severity assessments.

Conclusions:

• The developed system shows promise in accurately detecting VIMS in remote machinery operators.
• This research contributes to mitigating VIMS, thereby enhancing operator performance and safety in remote construction operations.
• Further development can facilitate wider adoption of remote-controlled construction machinery.