实时交通事件影响评估模型研究及其应用

项目名称： 实时交通事件影响评估模型研究及其应用

项目编号： No.51308559

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 建筑科学

项目作者： 钟任新

作者单位： 中山大学

项目金额： 25万元

中文摘要： 交通网络对交通事件非常脆弱。交通事件一旦发生，通常会快速形成道路通行能力的瓶颈。如不能及时处理，将产生严重的车辆拥挤排队现象，形成的交通拥堵可能会扩散并导致网络范围内的大面积交通拥堵。准确预估交通事件的事件影响范围、持续时间及其引发的行车延误，才能及时的制定有效的管理措施和交通转移策略等以疏导事件引发的交通拥堵。这些迫切性使得对交通事件引发的交通拥堵形成机理和传播规律以及拥堵的控制策略的研究成为国际性的热点研究课题。为了满足这些迫切的需求,本项目将：从多源数据出发,发展有效的数据滤波和融合技术为项目提供数据支撑；通过交通事件引入的不确定性、随机性的建模以研究交通事件引发的交通拥堵传播规律和消散机理；结合数据驱动的方法,使得上述研发的模型不但对数据的噪声等随机要素具备鲁棒性，并且对交通事件引发的突变有自适应性；最后,给上述的框架发展分布式和并行计算结构提高其运行效率,使得它能在大网络上应用。

中文关键词： 交通事件影响评估；随机交通流模型；假设驱动；数据驱动；数据滤波与融合

英文摘要： Traffic Incidents are non-recurring events that disrupt the normal traffic flow and induce motorist delay. Traffic incidents on density road network can result in gridlocks and severe congestion problems, which highlights the vulnerability of the road network to traffic incidents. This emerges transport departments around the world to investigate a possible introduction of a Traffic Incident Detection and Management System (TIDMS) to remedy this weakness of the road infrastructure. This proposal aims to contribute to this timely issue by developing a framework to assess the impacts of traffic incidents comprehensively. Key theoretical and practical challenges are: multiple data sources; spatial-temporal correlated observations; observations may be contaminated by noise induced by detector errors and imperfect communication channels, while traffic conditions during and after an incident often become abnormal and hardly predictable. In this project, we propose an online traffic incident impact assessment framework that integrates data (observation)-driven and assumption (model)-driven approaches under the umbrella of stochastic dynamic traffic flow modeling by fusing different traffic data sources with noise. The integration of data-driven and assumption-driven methods compensates the weakness of each method in th

英文关键词： traffic incident impact assessment；stochastic traffic flow model；assumption driven；data driven；data filtering and fusion