Elevated mental workload (MWL) experienced by pilots can result in increased reaction times or incorrect actions, potentially compromising flight safety. This study aims to develop a functional system to assist administrators in identifying and detecting pilots’ real-time MWL and evaluate its effectiveness using designed airfield traffic pattern tasks within a realistic flight simulator. The perceived MWL in various situations was assessed and labeled using NASA Task Load Index (NASA-TLX) scores. Physiological features were then extracted using a fast Fourier transformation with 2-s sliding time windows. Feature selection was conducted by comparing the results of the Kruskal-Wallis (K-W) test and Sequential Forward Floating Selection (SFFS). The results proved that the optimal input was all PSD features. Moreover, the study analyzed the effects of electroencephalography (EEG) features from distinct brain regions and PSD changes across different MWL levels to further assess the proposed system’s performance. A 10-fold cross-validation was performed on six classifiers, and the optimal accuracy of 87.57% was attained using a multi-class K-Nearest Neighbor (KNN) classifier for classifying different MWL levels. The findings indicate that the wireless headset-based system is reliable and feasible. Consequently, numerous wireless EEG device-based systems can be developed for application in diverse real-driving scenarios. Additionally, the current system contributes to future research on actual flight conditions.

Mental workload (MWL) recognition of carrier-based aircraft pilots benefits understanding the pilots' performance, which is crucial for enhancing the combat capability of carrier-based aircraft and ensuring flight safety. This paper aims to investigate the real-time MWL of pilots in carrier flight tasks, including takeoff, cruise, and landing. Firstly, we acquired the National Aeronautics and Space Administration-Task Load Index scales (NASA-TLX) and electroencephalogram (EEG) signals of 6 participants under different flight phases based on a flight simulator and a portable EEG. Secondly, the NASA-TLX scores were analyzed by statistical analysis. It is shown that the MWL levels are ranked as follows: landing > cruise > takeoff. Thirdly, the EEG features in the time domain, frequency domain, and nonlinear were extracted and these features were selected by statistical analysis and random forest feature importance evaluation. Finally, the performances of the k-nearest neighbor (KNN), random forest (RF), and support vector machine (SVM) were compared for the MWL identification. EEG Features with significant differences and high importance in different flight phases were employed as the inputs to three machine learning models. The results show that machine learning models can effectively identify pilots’ MWL, and the best identification of MWL was achieved using the selected fusion features as inputs to the RF classifier, with an average accuracy of 90.5%. The research results indicated that the portable EEG device is feasible for collecting high-quality EEG signals, and the RF classifier and selected fusion features are more suitable for identifying carrier-based pilots’ MWL.

Silent automation failures, where a system fails to detect a hazard without warning, pose a critical safety challenge for partially automated vehicles. While research has mostly focused on takeover requests, how to support a driver in silent failure remains underexplored. We conducted a multi-modal driving simulator study with 48 participants to investigate how different Prospective Situation Awareness Enhancement (PSAE) interfaces, delivered via augmented reality head-up display, affect takeover performance. By integrating behavioral, subjective psychological, and physiological data, our analysis suggests that situational awareness (SA) serves as an important moderating factor through which PSAE interfaces improve takeover performance. Further, we found that providing perceptual cues was most effective in enhancing SA, while communicating system intent was superior for building trust. Finally, we identified a potential correlate of SA in the neuroactivity. Overall, this paper contributes to understanding how transparency-oriented interfaces may support drivers and provides design insights into HMI design for silent failures.

Long-haul truck drivers face a high risk of drowsiness due to extended working hours and monotonous driving conditions. While Advanced Driver Assistance Systems (ADAS) are increasingly adopted, their impact on drowsiness accumulation in real-world heavy-truck operations remains unclear. Furthermore, the reliability of physiological and eye-tracking indicators for drowsiness monitoring in these specific contexts is not well established. Thus, a naturalistic driving study was conducted to analyze how operational timing (day vs. night starts) and driving mode (manual vs. ADAS-supported) influence driver drowsiness and the association between the drowsiness and physiological and eye-tracking indicators. Eight professional truck drivers participated in the study across 16 freight sessions on a 1,550 km route, yielding approximately 155 hours of valid data. Subjective drowsiness (KSS) was recorded hourly and synchronized with ocular, cardiovascular, and respiratory indicators. Mixed-effects models were employed to analyze baseline states, drowsiness trajectories, and indicator associations. Results indicate that night-start sessions began with higher drowsiness levels. Crucially, KSS increased over time during manual driving but remained stable under ADAS. Moreover, the association between physiological indicators (e.g., PERCLOS, RMSSD) and subjective drowsiness was found to be context-dependent, varying significantly with driving mode and operational timing. The study concludes that drowsiness monitoring in long-haul ADAS-supported freight operations must account for specific contextual factors, as standard physiological indicators may not generalize across different driving modes and session-start timings.