交通流预测是智能交通系统中的一类重要问题。尽管当前交通流预测方法取得了较好的进展,但还面临2个关键挑战:(1)交通流的变化模式不仅依赖于时间维度上的历史信息,还依赖于空间维度上相邻区域的信息,如何兼顾两个维度上的变化模式;(2)...交通流预测是智能交通系统中的一类重要问题。尽管当前交通流预测方法取得了较好的进展,但还面临2个关键挑战:(1)交通流的变化模式不仅依赖于时间维度上的历史信息,还依赖于空间维度上相邻区域的信息,如何兼顾两个维度上的变化模式;(2)时间本身具有小时、天及周等多粒度特性,如何实现多粒度下时序模式的捕捉。本文针对交通流预测的上述挑战,设计了一个多粒度时空深度回归模型(Spatial-temporal Deep Regression Model for Multi-granularity,MGSTDR),其基本思想是在多粒度时空交通流信息的基础上,对典型的差分整合移动平均自回归模型(Autoregressive Integrated Moving Average model,ARIMA)进行深度拓展,该模型在有效利用自身历史信息的同时,能兼顾相邻区域的信息,从而能够实现多粒度的时序交通流量预测。多个数据集上的实验结果表明,该模型在多粒度预测任务上优于现有的多个基准模型,尤其在小时这一粒度的预测结果上有5.66%的提升。展开更多
In order to improve the performance of the signalized intersection,an unconventional scheme tandem design(TD)is proposed.A simulation experiment is conducted to evaluate the capacity and delay under the unconventional...In order to improve the performance of the signalized intersection,an unconventional scheme tandem design(TD)is proposed.A simulation experiment is conducted to evaluate the capacity and delay under the unconventional scheme and two conventional lane assignment schemes.First,the VISSIM is employed as microsimulation to obtain the delay of different designs at signalized T-intersections under different conditions of traffic flow and turning proportion.Secondly,a method based on discriminant analysis(DA)is proposed to determine the best design scheme using the flow and turning proportion as inputs.Finally,a case study in Changsha city,China is used to demonstrate the efficiency and accuracy of these findings.The results indicate that the traffic flow and turning proportion are the crucial factors in scheme selection of lane assignment.Different from the previous research,the TD has better performance over various traffic flow levels.Furthermore,a proper proportion of left turns makes TD an outstanding option,which can reduce the delay and decrease the average number of stops and queue length significantly.However,the proportion should not be too high or too low.The research results can help practitioners obtain a quantitative view of appropriate design schemes at signalized intersections when trying to relieve traffic congestion according to different traffic conditions.展开更多
文摘交通流预测是智能交通系统中的一类重要问题。尽管当前交通流预测方法取得了较好的进展,但还面临2个关键挑战:(1)交通流的变化模式不仅依赖于时间维度上的历史信息,还依赖于空间维度上相邻区域的信息,如何兼顾两个维度上的变化模式;(2)时间本身具有小时、天及周等多粒度特性,如何实现多粒度下时序模式的捕捉。本文针对交通流预测的上述挑战,设计了一个多粒度时空深度回归模型(Spatial-temporal Deep Regression Model for Multi-granularity,MGSTDR),其基本思想是在多粒度时空交通流信息的基础上,对典型的差分整合移动平均自回归模型(Autoregressive Integrated Moving Average model,ARIMA)进行深度拓展,该模型在有效利用自身历史信息的同时,能兼顾相邻区域的信息,从而能够实现多粒度的时序交通流量预测。多个数据集上的实验结果表明,该模型在多粒度预测任务上优于现有的多个基准模型,尤其在小时这一粒度的预测结果上有5.66%的提升。
文摘In order to improve the performance of the signalized intersection,an unconventional scheme tandem design(TD)is proposed.A simulation experiment is conducted to evaluate the capacity and delay under the unconventional scheme and two conventional lane assignment schemes.First,the VISSIM is employed as microsimulation to obtain the delay of different designs at signalized T-intersections under different conditions of traffic flow and turning proportion.Secondly,a method based on discriminant analysis(DA)is proposed to determine the best design scheme using the flow and turning proportion as inputs.Finally,a case study in Changsha city,China is used to demonstrate the efficiency and accuracy of these findings.The results indicate that the traffic flow and turning proportion are the crucial factors in scheme selection of lane assignment.Different from the previous research,the TD has better performance over various traffic flow levels.Furthermore,a proper proportion of left turns makes TD an outstanding option,which can reduce the delay and decrease the average number of stops and queue length significantly.However,the proportion should not be too high or too low.The research results can help practitioners obtain a quantitative view of appropriate design schemes at signalized intersections when trying to relieve traffic congestion according to different traffic conditions.
