数据驱动的偶发拥堵时空建模及传播分析

为研究实际的道路交通路网中偶发拥堵的传播和演化特性，充分发挥海量交通流数据的潜在价值，克服现有基于模拟仿真的拥堵分析方法因理论假设和参数设置所导致的 “失真”问题，本文在交通流实测数据的基础上，建立改进的 PLS-STAR模型对偶发拥堵的时空传播结构进行描述，并提出偶发拥堵的直接和间接时空传播效应两种概念对拥堵的时空传播影响进行刻画，从而构造了一种数据驱动的偶发拥堵时空传播效应评估方法.通过北京路网的案例研究发现，路网服务水平的降低，更大程度来源于拥堵传播的间接影响而非直接取决于突发的交通量增加，因此，通过控制拥堵传播来提升城市路网的服务水平仍具有巨大潜力.

Data-driven Space-time Modeling and Propagation Study of Network-wide Sudden Congestion

For further analyzing the propagation and evolution characteristics of sudden congestion in actual road network, eliminating the distortion problem commonly existing in traditional simulation-based method because of superfluous assumptions or parameters, and taking advantage of the oceans of traffic flow data, an improved PLSSTAR model is developed to characterize the complex space-time propagation structure of sudden congestion on the basis of analyzing and identifying the sudden disruption using real traffic flow data. Then a congestion propagation effect evaluation method based on PLS-STAR is proposed based on the direct and indirect effects of space-time congestion propagation. By case studies on the road network in Beijing, it can be concluded that the reduction of road network’s service level is more attributed to the indirect effect of traffic congestion propagation, rather than the sudden increase of traffic volume. Therefore, there is still great potential to improve the service level of the urban road network by controlling congestion propagation effect.