Field observations illustrated that, right-turn vehicles stopped at various positions when proceeding within the right-turn lanes, while some of them trespassed on the crosswalks with multiple stops. In this case, ped...Field observations illustrated that, right-turn vehicles stopped at various positions when proceeding within the right-turn lanes, while some of them trespassed on the crosswalks with multiple stops. In this case, pedestrians and bikes (ped/bike) are encountered unsmooth and hazardous crossings when right-turn vehicles encroaching their lanes. Meanwhile, this also causes conflicts between right-turn and through vehicles at the crossing street. To better protect ped/bike at crossings with right-turn vehicles, this paper proposes a concept of “right-turn vehicle box” (RTVB) as a supplemental treatment within right-turn lanes. Sight distance, geometric conditions, and behaviors of vehicles and ped/bike are key factors to consider so as to set up the criteria and to design the suitable treatment. A case study was conducted at an intersection pair in Houston, USA to shape the idea of RTVB, together with driving simulator tests under relevant scenarios. The preliminary crosscheck examination shows that the right-turn vehicle box could possibly provide ped/ bike with smoother and safer crossings. In the interim, the safety and efficiency of right-turn operations were also improved. To further validate the effects, implementation studies should be conducted before the RTVB can make its debut in practice. Future works will focus on the complete warrants and design details of this treatment. Moreover, the concept of “vehicle box” could also be transplanted to other places where turning movement(s) needs assistance or improvements.展开更多
为及时识别、预测车辆的换道行为,综合考虑目标车辆及周边车辆的时空交互关系,结合时间卷积网络(Temporal Convolutional Network,TCN)的时序处理能力和长短期记忆(Long Short Term Memory,LSTM)神经网络的门控记忆机制,构建了基于TCNL...为及时识别、预测车辆的换道行为,综合考虑目标车辆及周边车辆的时空交互关系,结合时间卷积网络(Temporal Convolutional Network,TCN)的时序处理能力和长短期记忆(Long Short Term Memory,LSTM)神经网络的门控记忆机制,构建了基于TCNLSTM网络的车辆换道意图识别模型。首先,将目标车辆的驾驶意图分为直行、向左换道和向右换道3种类型,从CitySim车辆轨迹数据集中提取出目标车辆及对应同车道、左侧车道、右侧车道的相邻前车和相邻后车的轨迹数据,并利用中值滤波算法获得车辆运行状态指标。其次,针对统计学理论和机器学习方法面临的识别精度不高、训练时间长、参数更新慢等问题,提出利用膨胀卷积技术提取时间序列的时序特征,采用门控记忆单元捕捉时序特征的长期依赖关系,并以目标车辆及周边相邻车辆的速度、加速度、航向角、航向角变化率和相对位置信息等54个车辆状态指标为输入变量,以车辆的换道意图为输出变量,构建了一个基于TCN-LSTM网络的车辆换道意图识别模型。最后,对比分析了不同输入时间步长下TCN、支持向量机(Support Vector Machines,SVM)、LSTM和TCN-LSTM模型的识别精度。结果表明:输入时间序列长度为150帧时,TCN-LSTM模型的识别精度达到最高值96.67%;从整体分类精度来看,相比LSTM、TCN和SVM模型,TCN-LSTM模型的换道意图分类准确率分别提升了1.34、0.84和2.46个百分点,展现出了更高的分类性能。展开更多
Through and right-turn shared lanes are widely designed to increase the capacity of through traffic,but they can also cause delays for right-turn vehicles.This study presents a dynamic control method for a shared lane...Through and right-turn shared lanes are widely designed to increase the capacity of through traffic,but they can also cause delays for right-turn vehicles.This study presents a dynamic control method for a shared lane that prioritizes right-turn vehicles at the beginning of the cycle and subsequently allows through traffic to queue in the shared lane for saturated discharge.The traffic wave model is employed to reveal the dynamics of the traffic flow under this control and to derive the relationships among major traffic parameters.Constrained by the major relationship,a linear programming approach to minimize the total queue length is developed to determine the proper values of control parameters,including the shared area length,subordinate signal time lag,and shared or exclusive duration.A sensitivity analysis of the control parameters for different arrival rates and flow ratios is performed.Comparisons are conducted among the dynamic shared lane,the fixed exclusive lane,and the fixed shared lane.The results show that the dynamic control method results in a lower delay for both through and total traffic.展开更多
文摘Field observations illustrated that, right-turn vehicles stopped at various positions when proceeding within the right-turn lanes, while some of them trespassed on the crosswalks with multiple stops. In this case, pedestrians and bikes (ped/bike) are encountered unsmooth and hazardous crossings when right-turn vehicles encroaching their lanes. Meanwhile, this also causes conflicts between right-turn and through vehicles at the crossing street. To better protect ped/bike at crossings with right-turn vehicles, this paper proposes a concept of “right-turn vehicle box” (RTVB) as a supplemental treatment within right-turn lanes. Sight distance, geometric conditions, and behaviors of vehicles and ped/bike are key factors to consider so as to set up the criteria and to design the suitable treatment. A case study was conducted at an intersection pair in Houston, USA to shape the idea of RTVB, together with driving simulator tests under relevant scenarios. The preliminary crosscheck examination shows that the right-turn vehicle box could possibly provide ped/ bike with smoother and safer crossings. In the interim, the safety and efficiency of right-turn operations were also improved. To further validate the effects, implementation studies should be conducted before the RTVB can make its debut in practice. Future works will focus on the complete warrants and design details of this treatment. Moreover, the concept of “vehicle box” could also be transplanted to other places where turning movement(s) needs assistance or improvements.
文摘为及时识别、预测车辆的换道行为,综合考虑目标车辆及周边车辆的时空交互关系,结合时间卷积网络(Temporal Convolutional Network,TCN)的时序处理能力和长短期记忆(Long Short Term Memory,LSTM)神经网络的门控记忆机制,构建了基于TCNLSTM网络的车辆换道意图识别模型。首先,将目标车辆的驾驶意图分为直行、向左换道和向右换道3种类型,从CitySim车辆轨迹数据集中提取出目标车辆及对应同车道、左侧车道、右侧车道的相邻前车和相邻后车的轨迹数据,并利用中值滤波算法获得车辆运行状态指标。其次,针对统计学理论和机器学习方法面临的识别精度不高、训练时间长、参数更新慢等问题,提出利用膨胀卷积技术提取时间序列的时序特征,采用门控记忆单元捕捉时序特征的长期依赖关系,并以目标车辆及周边相邻车辆的速度、加速度、航向角、航向角变化率和相对位置信息等54个车辆状态指标为输入变量,以车辆的换道意图为输出变量,构建了一个基于TCN-LSTM网络的车辆换道意图识别模型。最后,对比分析了不同输入时间步长下TCN、支持向量机(Support Vector Machines,SVM)、LSTM和TCN-LSTM模型的识别精度。结果表明:输入时间序列长度为150帧时,TCN-LSTM模型的识别精度达到最高值96.67%;从整体分类精度来看,相比LSTM、TCN和SVM模型,TCN-LSTM模型的换道意图分类准确率分别提升了1.34、0.84和2.46个百分点,展现出了更高的分类性能。
基金supported by the National Natural Science Foundation of China(Nos.52325210,52131204,and 52302411)the Shanghai Science and Technology Innovation Action Plan Project(No.19DZ1209004)the Fundamental Research Funds for the Central Universities(No.22120220137).
文摘Through and right-turn shared lanes are widely designed to increase the capacity of through traffic,but they can also cause delays for right-turn vehicles.This study presents a dynamic control method for a shared lane that prioritizes right-turn vehicles at the beginning of the cycle and subsequently allows through traffic to queue in the shared lane for saturated discharge.The traffic wave model is employed to reveal the dynamics of the traffic flow under this control and to derive the relationships among major traffic parameters.Constrained by the major relationship,a linear programming approach to minimize the total queue length is developed to determine the proper values of control parameters,including the shared area length,subordinate signal time lag,and shared or exclusive duration.A sensitivity analysis of the control parameters for different arrival rates and flow ratios is performed.Comparisons are conducted among the dynamic shared lane,the fixed exclusive lane,and the fixed shared lane.The results show that the dynamic control method results in a lower delay for both through and total traffic.
