基于驾驶行为实验的城市快速路主动安全速度管理研究

Traffic safety has become a severe issue for the urban expressway system, which has greatly influenced the operation efficiency. As been proved by several international metropolitans, proactive safety speed management could be an appropriate approach to improve the expressway operation safety. Two key issues for proactive safety speed management are management timing prediction and optimal management strategies identification. Current researches were mainly conducted at the traffic operation level, where the relationships between crash and traffic operation features were established and the operation statuses were predicted relying on the aggregate traffic flow model (fundamental diagram). However, traffic crashes were mainly attributed by driving behaviors while traffic operation characteristic is an aggregation of individual behaviors. Therefore, it is essential to consider human factor issues when developing the management strategies. ..In this study, driving behavior experiment data will be collected through naturalistic driving study (NDS) and advanced driving simulator platforms. Both crash risk evaluation and causal inference models will be developed for the safety analysis aspects; while for traffic flow prediction, driving behavior characteristics will be accounted. Finally, human factor oriented proactive safety speed management strategies will be established. More specifically, the causal relationships between geometric characteristics, traffic operation, driving behavior and traffic crash will be identified based on the NDS data. Individual driving behavior will be analyzed through driving simulator and cooperative driving simulation platform. Deep learning technique will be applied to establish the real-time crash risk evaluation model. Based on the driving behavior analysis model, macroscopic traffic flow model would be established while accounting for the human factor. Finally, under the Model Predictive Control (MPC) scheme, proactive safety speed management strategies will be established and the influencing factors will be identified. Through systematic investigation, the formulated fundamental theories and key techniques will provide supportive for future applications for the proactive safety speed management in China.

我国城市快速路交通事故频发，对其运行效率影响大；需借鉴发达国家经验开展主动安全速度管理。主动安全速度管理的核心在于实施时机预判和管理策略优选：目前相关研究局限于交通运行层面，基于事故数据进行风险评估，通过理论假设开展运行预测；但事故致因由人为因素占主导，运行状态则是驾驶人行为的集计，需在驾驶行为层面开展机理研究。本课题拟通过自然驾驶与驾驶模拟相结合的行为实验、风险评估与机理推断相结合的安全分析、行为建模与态势演化相结合的运行预测，构建“人因导向”的主动安全速度管理策略。基于自然驾驶实验，从交通运行、驾驶行为等角度探索事故致因；基于驾驶模拟实验，揭示驾驶人与速度管理的交互作用机理；构建全样本数据环境下的事故风险评估方法；建立考虑人为因素的宏观交通流预测模型。最后，基于模型预测控制框架，构建速度管理策略并开展安全改善效果评估。通过研究，为主动安全速度管理在我国城市快速路的应用提供有力支撑。

我国城市快速路交通事故频发，对其运行效率影响大，亟需开展主动安全速度管理，调控风险交通状态。主动安全速度管理的核心在于实施时机预判和管理策略优选：目前相关研究局限于交通运行层面，基于事故数据进行风险评估，通过理论假设开展运行预测；但事故致因由人为因素占主导，运行状态则是驾驶人行为的集计，因此需在驾驶行为层面开展机理研究。.本项目通过自然驾驶与驾驶模拟相结合的行为实验、风险评估与机理推断相结合的安全分析、行为建模与态势演化相结合的运行预测，构建了“人因导向”的主动安全速度管理策略，完成了“事故致因解析-行为机理分析-事故风险辨识-主动管控策略优化”等核心方法与关键技术的研究任务。具体研究内容包括：1）基于自然驾驶实验的快速路事故致因分析方法; 2）基于驾驶模拟实验的驾驶人反应行为机理分析; 3）主动安全速度管理事故风险评估方法; 4）主动安全速度管理运行态势预测模型; 5）主动安全速度管理策略构建与效果分析。.本项目突破了事故致因同质假设，提出了考虑事故致因异质性的耦合影响因素关系解析方法；提出了异质性驾驶人反应行为机理的分析方法，构建了差异化驾驶能力的量化评估模型；提出了零膨胀、非平衡数据的事故风险量化评估方法；量化了交通运行态势与事故风险的关联关系；开展了事故风险预警阈值优选方法研究，构建了协同式管控策略优化方法。.依托本项目累计发表高水平学术论文15篇，包括Analytic Methods in Accident Research（JCR交通学科类别排名1/37）、Transportation Research Part C（JCR交通学科类别排名3/37）、Accident Analysis & Prevention（中科院一区）等高水平期刊论文。授权专利1项、做特邀学术报告3次。.项目成果支撑了上海快速路等重要基础设施的安全高效运行，2020年获上海市科技进步二等奖1项（排名第三）。