基于混合聚类的城市道路多时段控制研究

Mixed clustering based time of day control study in urban road

多方案控制具有简单实用、可靠性好、稳定性高、性价比高等优点,在城市交通控制中得到了广泛应用。在利用指数平滑异同移动平均法对缺失交通数据进行修补的基础上,运用快速聚类与系统聚类相结合的混合聚类方法对交通数据进行分析,并以经典的Silhouttte评价指标作为聚类终止条件,同时兼顾/考虑了交通数据时序性,消除聚类结果中的三类奇异点,最后利用交通信号配时软件SYNCHRO制定了相应时段的控制方案。相关的案例分析显示,优化后的多时段控制方案可以更好地适应交通流的波动,车均延误减少9.79%,其中时序性考虑可以有效改善控制方案频繁切换对交通流的扰动,避免其负效应对混合聚类改善效果的影响。

Time of day control has been widely applied in the urban traffic control, considering its simpleness, high reliabili- ty, and well cost benefit. With the method of moving average convergence divergence, this paper first mended the lost histori- cal data, and then it used mixed clustering in traffic data analysis, which combined K-means clustering with system clustering, and used classic Silhouttte index as the termination condition. Meanwhile, in consideration of time sequence of traffic flow, eliminated singular points in clustering, which classified in 3 classes. After that, it used traffic signal timing optimization soft- ware SYNCHRO to make the corresponding control plans in different intervals. The results indicate that the average decrement rate of average vehicle delay of the method is 9.79%, in which the optimal TOD control plans can also reduce the fluctuation during TOD transitions with the consideration of time sequence, thus avoid its negative effects on improvement of the mixed clustering method.