城市道路交通拥堵状态时空相关性分析0

摘要

The formation and dissipation processes of urban traffic congestions are influenced by shocking waves of different cycles inside the traffic flow. The original factors that lead to traffic congestions are very complicated, the modelling is therefore difficult. This is the main reason that research works about the spatio-temporal dissipation effects for congestions are eventually stopped at the level of qualitative analysis. Some quantitative analysis can be successfully done based on the measured traffic data. However, rare effective knowledge extraction methods can be found to deal with data containing information about multiple time scales and granularities, which however is important to correlation analysis and the direct use of original data leads to unstationary signal features and opposite observation conclusions when putting the data into the discussion of given time scales. Focused on the analysis of spatio-temporal correlation of traffic parameters in congestion areas, a new analyzing method is developed and used based on Pearson's correlation index, which decomposes a measured road speed trajectory into trend and detail components in different time scales. The initial verification and application of this method and the corresponding data segmentation algorithm show the quantitative characteristics of the congestion diffusion in time and space by observing the variation of correlation status.%城市交通拥堵的形成和消散实际受多周期的交通流冲击波影响，产生来源非常复杂，建模也很困难。正因如此，有关拥堵时空扩散性的研究多停留在定性层面。基于直接采集交通数据可进行定量分析，但数据的细粒度特性使信号不够平稳，在多时间尺度上甚至表现出相反特征，缺乏有效的知识挖掘思路和方法。针对拥堵区域交通参数时空相关性问题，在皮尔逊相关性指标的基础上提出并采用了一种新的分析方法，它将道路实测速度轨迹分解为多时间尺度的趋势和细节分量。应用该方法提出的相关性指标和数据分段算法，对典型交通拥堵扩散算例进行了分析，借助相关性状态变化刻画了拥堵在时空中扩散的定量特征。