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
Due to the quick development of urbanization,it is important to provide a healthy urban environment for the dweller.Previous studies have obtained valuable conclusions of how to improve the urban airflow distribution ...Due to the quick development of urbanization,it is important to provide a healthy urban environment for the dweller.Previous studies have obtained valuable conclusions of how to improve the urban airflow distribution under isothermal conditions.How to adopt and interpret those conclusions when considering the solar-induced atmospheric stability conditions have not been clarified yet.In this study,the characteristics of atmospheric stability condition and influence of diurnal varying solar-induced thermal effect on urban airflow inside the idealized building arrays were investigated at five cities located at five climate zones in China.Urban energy model,CitySim,was employed to simulate the annual distribution of solar-induced walls’temperatures inside the idealized building arrays.The diurnal varying wall temperatures at the hottest days were set as thermal boundary conditions in computational fluid dynamic(CFD)simulations.With albedo value of 0.5,the possibility of adopting the results from isothermal condition directly ranged from 7%to 11%for the five cities in China throughout the year.The unstable condition ocuppied from 19%to 24%annually and the stable condition of more than 40%annually was observed.Under the diurnal varying solar-induced thermal effect,the spatially-averaged air speeds and airflow patterns were significantly different from the isothermal conditions.The percent of Richardson number under different atmospheric stability conditions annually based on the Citysim simulation results indicated that the atmospheric stability was most likely determined by the local climate characteristics and albedo value rather than the building layouts at the five selected cities in China,but this should be further investigated when the shadow effects of surrounding buildings were considered in simulations.展开更多
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The authors acknowledge the scholarship(JY201704)from Guangzhou Elite Project(GEP)granted for Guoxing Chen’s Ph.D.study at Aarhus University.
文摘Due to the quick development of urbanization,it is important to provide a healthy urban environment for the dweller.Previous studies have obtained valuable conclusions of how to improve the urban airflow distribution under isothermal conditions.How to adopt and interpret those conclusions when considering the solar-induced atmospheric stability conditions have not been clarified yet.In this study,the characteristics of atmospheric stability condition and influence of diurnal varying solar-induced thermal effect on urban airflow inside the idealized building arrays were investigated at five cities located at five climate zones in China.Urban energy model,CitySim,was employed to simulate the annual distribution of solar-induced walls’temperatures inside the idealized building arrays.The diurnal varying wall temperatures at the hottest days were set as thermal boundary conditions in computational fluid dynamic(CFD)simulations.With albedo value of 0.5,the possibility of adopting the results from isothermal condition directly ranged from 7%to 11%for the five cities in China throughout the year.The unstable condition ocuppied from 19%to 24%annually and the stable condition of more than 40%annually was observed.Under the diurnal varying solar-induced thermal effect,the spatially-averaged air speeds and airflow patterns were significantly different from the isothermal conditions.The percent of Richardson number under different atmospheric stability conditions annually based on the Citysim simulation results indicated that the atmospheric stability was most likely determined by the local climate characteristics and albedo value rather than the building layouts at the five selected cities in China,but this should be further investigated when the shadow effects of surrounding buildings were considered in simulations.
