| 其他摘要 | Road safety and fatigue driving have always been important issues to the government and the public. Research showed that 90% of road accidents were caused by individual drivers. And driving fatigue is one of the main influencing factors. Fatigue causes 20% of deaths and serious accidents. Therefore, it is necessary to thoroughly explore the physiological, psychological, and behavioral characteristics of the drivers after fatigue, and try to detect and predict the fatigue state, and provide advice for the intervention of fatigue driving, thus improving driving safety in the highest degree.
Study 1 focused on the impact of fatigue on the driver’s cognitive process, including perception processing and decision-making processing. This study included two experiments (Experiment 1 and 2) which aimed to explore the impact of fatigue on the driver’s estimation of different speeds and risk-taking in decision-making task respectively. Experiments 1a and 2a recruited 26 drivers. Participants need to come to the laboratory twice, one time to complete the speed estimation task and the risk decision task related to the driving situation (namely, BART) after fatigue (induced by long-term driving), the other time to complete the same task in the awake state. The order of fatigue and awake state is balanced among the subjects. The results of experiment 1a found that the estimated deviation was reduced after several training feedbacks when the driver was awake, but in the fatigue state there was no significant difference, indicating the learning effect of the participants on the feedback in the fatigue state decline. The procedures of Experiments 1b and 2b were the same as Experiments 1a and 2a, recruiting 32 drivers, increasing the duration of fatigue induction, and making task adjustments based on the results of Experiments 1a and 2a. The results of experiment 1b showed that the participants’ estimated deviations of slower cars were smaller than those of medium-speed cars when they were awake, but there was no difference in estimation deviations when fatigued, indicating that the position estimation bias of slower speed vehicles increases. The results of experiment 2b showed that the number of pumping-ups after adjustment in the fatigue state was significantly lower than that in the awake state, that is, the choice was more conservative.
Study 2 focused on the impact of fatigue on driver's driving behavior and its early warning, exploring the use of EEG signals to predict fatigue driving events in advance. 18 drivers were recruited and asked to maintain the current lane to drive for a long time, and report subjective fatigue every 5 minutes. EEG signals are recorded at the same time during driving. The results showed that horizontal displacement standard deviation and the longitudinal velocity standard deviation have an inverted U-shaped trend over time, which is significantly related to the subjective report change trend. Combining the subjective fatigue and driving control deviation (horizontal or longitudinal) of the subjects, and taking 5 minutes as the unit, three types of fatigue stages are extracted: 1) Awakening stage: Subjects are subjectively awake and driving control deviation is small, 2) Moderate fatigue stage: Subjectively fatigued and driving control deviation is small, 3) Severe fatigue stage: Subjectively fatigued and driving deviation is large. Extract the EEG data of the 5 minutes prior of each stage, and calculate the power values of the four frequency bands (alpha, beta, theta, delta) of the EEG. Use Logistic regression model for modeling. The results showed that the EEG signal of the previous stage could predict the fatigue state of the next stage, and the prediction accuracy could reach about 60%; and the prediction effect of the EEG signal was better than driving indicators (steering wheel angle or accelerator acceleration).
In general, this study found that the cognitive resources of the driver decreased after fatigue, thus the estimation bias for slower speed cars increased, and the risk-taking tendency decreased in a risky decision-making task. In addition, the study found that the EEG signals could predict the adverse performance of lateral and longitudinal control induced by fatigue in the next stage, and the use of Logistic regression modeling to predict the three classification models of awake, moderate fatigue, and severe fatigue could reach an accuracy in approximately 60%. This research further supplements and expands the discussion of the impact of driving fatigue on the cognitive process and the prediction of fatigue driving. In addition, the research results can provide guidance for the detection and early warning of fatigue driving. |
修改评论