Hazard perception (HP) is the process of detecting and identifying hazards. Drivers’ HP abilities are critical for driving safety. This paper presents a systematic literature review of driver HP, including scientific measures of HP, major human factors affecting HP, and training methods for improving HP skills. Sixty-nine peer-reviewed studies were identified and reviewed. The results showed that common measures of HP include hazard scenario questionnaires, HP reaction time, hazard hit rate, and eye fixation measures such as fixation probability, fixation reaction time, fixation duration, and fixation variance. Major human factors that affect HP include experience, aging, fatigue, distraction, and the use of alcohol and drugs. Various training methods have been developed to train and improve drivers’ HP skills. In general, there is evidence in the literature showing the effectiveness of HP training. A combination of complementary training approaches such as instruction, expert demonstration, and active practice with feedback and attention support the use of picture-, video-, computer-, and simulator-based training methods to improve HP performance in shorter HP reaction time, higher hazard hit rate, and better eye scan patterns (more spread scan, more anticipatory scan). These findings could guide future work developing and designing HP training programs. Three future research areas are identified and discussed: the need for standardized HP tests, long-term testing of HP training programs, and new HP questions and challenges brought by partially automated vehicles.

Young drivers are overrepresented in motor vehicle crashes, and are shown to be poorer at anticipating potential threats on the roadway compared with their more experienced peers. Literature demonstrates the effectiveness of driver training programs at improving young drivers’ latent hazard anticipation performance. Various hazard anticipation training studies have been undertaken on different population demographics using different training scenario presentation modes and multiple evaluation testbeds. These error-based feedback training programs (3M) allow trainees to make a mistake, show them how to mediate the mistake, and provide an opportunity to master the target skills. The current meta-analytical review focused on 19 peer-reviewed training studies that utilized eye movements to measure improvements in drivers’ latent hazard anticipation performance following training. The role of four moderating factors (mode of delivery – PC-based or non PC-based; presentation of training – egocentric or exocentric; method of evaluation – on-road or driving simulator; and age of sample – teen novices aged 16–17 or young drivers aged 18–21) on the training effects were explored. Overall, the current meta-analysis suggest that: (a) superficial improvements in training programs does not necessarily further improve the drivers’ latent hazard anticipation; (b) drivers who completed a training program with both egocentric and exocentric training views achieved greater levels of latent hazard anticipation performance than those who completed a training program that contained either view, but not both; and (c) the effect sizes of the 3M-based training programs on latent hazard anticipation were greater for drivers aged 18–21 years than drivers aged 16–17.

This study investigated whole-body vibration (WBV) response in real field harrowing operations at different tractor ride conditions i.e., average speed, front harrow pin angle (FHPA), and rear harrow pin distance (RHPD). Taguchi’s L₂₇ orthogonal array was used to formulate a systematic design of experiments. WBV exposure was measured along the three translational axes to compute overall daily vibration magnitude i.e., A(8). Tractor’s seat isolation capacity was assessed in terms of Seat Effective Amplitude Transmissibility i.e., SEAT%. Raw acceleration data was analysed to obtain dominant frequencies using Fast Fourier Transform (FFT). A(8) was found to range between 0.43 to 0.87 m/s² in the experimental trials. Seat isolation capacity was found to be poor in 89% of the experiments i.e., SEAT% > 100%. Average speed and FHPA was found to have a significant impact (p ≤ 0.05) on A(8) and SEAT%. FFT response showed a range of primary and secondary dominant peaks within a frequency range of 0.2 to 11 Hz. In conclusion, the majority of experimental trials (67%) exceeded the Directive2002/44EU recommended exposure action value (EAV) limit i.e., 0.5 m/s². The harrowing operation was found to exhibit vibration exposure at low frequencies in the vicinity of natural frequencies of the human body and may consequently affect ride comfort.

Effective driver re-engagement is essential for the safe operation of automated driving systems (ADS), especially during takeover requests. This study examines the effects of alert intensity and driver engagement in non-driving related tasks on takeover performance in automated vehicles. Forty-one participants navigated simulated driving scenarios with various hazards and alert conditions, allowing analysis of response times, physiological responses, and subjective perceptions. The findings showed that higher-intensity alerts significantly reduced reaction times, leading to quicker driver takeovers. Interestingly, engagement in certain secondary tasks sometimes improved driver responsiveness, suggesting that moderate cognitive engagement may enhance alertness. Response times varied across different driving scenarios, indicating the influence of situational context on driver behavior. Physiological measures, including eye-tracking and heart rate monitoring, showed increased cognitive and physiological arousal during takeover events, particularly in response to stronger alerts. Participants reported higher confidence and satisfaction with high-intensity alerts, without reporting increased annoyance, indicating a preference for more assertive alert mechanisms. These outcomes highlight the critical role of adaptive alert mechanisms in ADS design, promoting for context-aware signaling and real-time driver state monitoring to improve takeover performance and ensure vehicular safety. The study suggests that future ADS should integrate dynamic alert systems capable of adjusting intensity based on situational urgency and driver engagement, thereby enhancing overall system reliability and user acceptance.

Autonomous vehicles (AVs) face critical decisions in pedestrian interactions, necessitating ethical considerations such as minimizing harm and prioritizing human life. This study investigates machine learning models to predict human decision making in simulated driving scenarios under varying pedestrian configurations and time constraints. Data were collected from 204 participants across 12 unique simulated driving scenarios, categorized into young (24.7 ± 3.5 years, 38 males, 64 females) and older (71.0 ± 5.7 years, 59 males, 43 females) age groups. Participants’ binary decisions to maintain or change lanes were recorded. Traditional logistic regression models exhibited high precision but consistently low recall, struggling to identify true positive instances requiring intervention. In contrast, the AdaBoost algorithm demonstrated superior accuracy and discriminatory power. Confusion matrix analysis revealed AdaBoost’s ability to achieve high true positive rates (up to 96%) while effectively managing false positives and negatives, even under 1 s time constraints. Learning curve analysis confirmed robust learning without overfitting. AdaBoost consistently outperformed logistic regression, with AUC-ROC values ranging from 0.82 to 0.96. It exhibited strong generalization, with validation accuracy approaching 0.8, underscoring its potential for reliable real-world AV deployment. By consistently identifying critical instances while minimizing errors, AdaBoost can prioritize human safety and align with ethical frameworks essential for responsible AV adoption.