|
|
- 2016
融合先验经验聚类的终端区交通流相态识别
|
Abstract:
以终端区交通流为研究对象,基于航迹谱聚类结果定义并提取交通流特征,分析了特征间关系与交通流相态演化规律,发掘了实测数据下交通流的自由态、平稳态与拥堵态,以此为先验经验进一步设计因子分析与遗传期望最大化模糊聚类算法相结合的终端区交通流态势识别方法,实现对交通流状态影响因素与交通流隐性特征的提取,选取典型繁忙终端区的实测数据进行验证。分析结果表明:基于客观数据挖掘的交通流态势识别方法具有良好的适应性与准确性,自由态、平稳态与拥堵态的模型识别数量分别为6、36、37,管制员判别数量分别为7、40、32,误差率分别为14.3%、10.0%、15.6%,模型识别率均在84%以上; 提取的交通流相态及时空特征可从局部细节构建终端区整体运行态势,为终端区流量时空分布调配与进离场程序优化提供支撑。
The traffic flow in terminal area was taken as research object, and the characteristics of traffic flow were defined and extracted based on the result of trajectory spectral clustering. The relationship of characteristics and phase-state transition law of traffic flow were analyzed to reveal three phase-states of traffic flow under observed data, including free state, steady state and congestion state, which was regarded as prior experience to further design the identification method of traffic flow situation in terminal area combining factor analysis and fuzzy clustering algorithm of genetic expectation maximization, the influence factor of traffic flow state and the recessive characteristics of traffic flow were extracted, and the observed data from typical busy terminal area were chosen to do the verification. Analysis result shows that the identification method of traffic flow situation based on objective data mining has good adaptability and accuracy, the identification numbers by the method for free state, steady state and congestion state are 6, 36 and 37 respectively, the discrimination numbers by the controller are 7, 40 and 32 respectively, the error rates are 14.3%, 10.0% and 15.6% respectively, and the identification rates are all above 84%; the extracted phase-state and time-spatial characteristic of traffic flow can be used to structure the overall operation situation in terminal area from local detail, which can provide support for the time-spatial distribution allocation of flow in terminal area and the optimization of arrival and departure procedure. 3 tabs, 11 figs, 25 refs
