事件相关的驾驶压力对驾驶绩效有着负面影响，威胁驾驶员的身心健康。车辆智能技术与设备的发展为解决驾驶员状态感知与干预提供了可能性，其中包括对于驾驶压力的监测以及相应的干预。为此，本文在已有研究的基础上，提出从事件层面进行压力监测和即时干预，综合使用模拟驾驶和自然驾驶研究方法，开展了三项实验研究。 第一项研究聚焦于模拟驾驶环境下的压力诱发与监测，通过新设计的压力诱发方法来获得参试者相应的压力状态，在此过程中采集行为、生理和主观数据来训练机器学习模型。结果表明，该方法成功诱发了事件相关的驾驶压力，基于面部表情数据和K近邻算法的模型在事件层面取得89.2%的压力分类准确率，显著高于生理数据训练的模型。此外，加入个体特征可进一步提高模型的性能表现，还找到了面部表情中的压力敏感特征。 第二项研究聚焦于实路环境下的压力数据采集、监测与分析。从实验室环境转向实路环境，旨在提高监测方法的应用性。面对自然驾驶研究的挑战，本部分基于移动设备新设计了一个对驾驶任务无干扰的数据采集汇报系统，并通过自然驾驶实验检验了可行性。使用融合车辆运动学、生理和面部表情的数据和XGBoost算法开发的实路压力监测模型，取得了92.5%的准确率，面部表情数据训练的模型优于其他的单一数据类型，加入个体特征可提高模型性能。 第三项研究聚焦于驾驶情境下的压力即时干预。在准确监测驾驶压力后，本部分从驾驶绩效、生理数据和主观汇报三个角度评估了两种即时压力干预措施（仅语音正向评论和额外加入基于触觉的呼吸引导）的有效性。结果表明，与基线相比参与干预并不会显著影响驾驶绩效，而无干预组的车辆控制表现则会更差。触觉呼吸引导可以有效降低呼吸速率至目标值。在简单路况下，呼吸干预显著提高了RMSSD，但在复杂路况下并没有带来健康帮助却显著降低了主观评价，因此干预措施的选择应当考虑路况复杂程度。 本文提出的方法是对已有驾驶压力研究的拓展和深入，可以促进相关研究迁移至实路环境，提高研究的应用价值。研究成果将帮助实现驾驶员实时状态监测，提高车辆智能化水平，便于驾驶员自我压力管理，最终提高驾驶的安全性和舒适性。

Event-related driver stress harms driving performance and threatens drivers‘ physical and mental health. The development of intelligent vehicle technology and devices makes it possible to monitor and intervene in driver status, including driver stress monitoring and intervention. This dissertation proposed stress monitoring and immediate interventions at the event level and conducted three experimental studies using a combination of simulated and naturalistic driving research methods. The first study focused on stress elicitation and monitoring in a simulated driver environment. A newly designed stress elicitation method was used to result in the required stress status of the participants. Participants‘ behavioral, physiological, and subjective data were collected to train machine learning models. The results showed that the method successfully induced event-related driver stress, and the model based on facial expression data and the K-nearest neighbor algorithm achieved 89.2\% stress classification accuracy at the event level, which was significantly higher than that of the model trained with physiological data. In addition, adding individual features can further improve the performance of the model. We also found stress-sensitive features in facial expressions. The second study focused on stress data acquisition, monitoring, and analysis in a real road environment. The shift from the laboratory environment to the real-road environment aimed to improve the applicability of existing monitoring methods. Facing the challenges of naturalistic driving studies, a new data acquisition and reporting system based on mobile devices without interference to driving tasks was designed in this part, and the feasibility was tested by naturalistic driving experiments. A real road stress monitoring model developed using a fusion of vehicle kinematic, physiological, and facial expression data and the XGBoost algorithm achieved an accuracy of 92.5\%. Models trained on facial expression data outperformed other single data types, and including individual features improved model performance. The third study focused on immediate intervention of driver stress in driving situations. After accurately monitoring driver stress, this section evaluated the effectiveness of two immediate stress interventions (positive verbal comments and additional inclusion of tactile-based breathing guidance) from three perspectives: driving performance, physiological data, and subjective evaluation. Results indicated that engaging in the intervention did not significantly affect driving performance compared to baseline, whereas vehicle control performance was worse in the no-intervention group. Tactile breathing guidance effectively regulated breathing rate to target values, and the breathing intervention significantly improved RMSSD in simple road conditions but did not provide health benefits but significantly reduced subjective ratings in complex road conditions. Therefore, the choice of interventions should consider the complexity of the road conditions.The method proposed in this dissertation is an extension and deepening of the existing research on driver stress, which can facilitate the migration of related research to the real road environment and increase the application value. The research results will help realize driver status monitoring, improve vehicle intelligence, facilitate driver stress management, and improve driving safety and comfort.