The successful application of new technologies such as remotely piloted aircraft systems,distributed electric propulsion systems,and automatic control systems on electric vertical take-off and landing(eVTOL)aircraft h...The successful application of new technologies such as remotely piloted aircraft systems,distributed electric propulsion systems,and automatic control systems on electric vertical take-off and landing(eVTOL)aircraft has prompted Urban Air Mobility(UAM)to be mentioned frequently.UAM is a newly raised transport mode of using eVTOL aircraft to transport people and cargo in urban areas,which is thought to share some of the traffic on the ground.One of the prerequisites for UAM to operate on a regular basis is that its demand can support the operating costs,so forecasting UAM demand is necessary.We conduct UAM demand forecasting based on the four-step method,focusing on improving the third-step modal split,and propose a demand forecasting model based on the logit model.The model combines a nested logit(NL)model with a multinomial logit(MNL)model to solve the problem of non-existent UAM sharing rates.We use Chengdu,China as an example,and focus on forecasting the UAM traffic demand in 2030 with the help of the four-step method.The results show that UAM is suitable for shared operation during the early stages.With a fully shared operation,the UAM share rate increases by 0.73%for every kilometer increase in distance.Moreover,UAM is more competitive than other modes for delivery distances exceeding 15 km.Finally,using the distributions of the share rate and traffic flow pattern from the simulation,we propose the routes that can be prioritized for UAM operations in Chengdu.展开更多
基于宁波市公共自行车刷卡数据、POI(Point of Interest)数据、气象和空气质量等数据,从数据驱动视角,深入挖掘公共自行车使用的时空特征及站点租还车需求预测。在时间上,采用KMeans算法,将站点聚为5类,探讨各类站点的时变需求规律及影...基于宁波市公共自行车刷卡数据、POI(Point of Interest)数据、气象和空气质量等数据,从数据驱动视角,深入挖掘公共自行车使用的时空特征及站点租还车需求预测。在时间上,采用KMeans算法,将站点聚为5类,探讨各类站点的时变需求规律及影响因素;在空间上,提出基于POI数据的站点用地类型识别方法,将站点分为居住类、交通设施类、办公类和商业休闲类。构建以15,30,60 min为间隔,以租还车需求为目标变量的随机森林预测模型,并与常用的BP(Back Propagation)神经网络、K最近邻方法进行比较。结果表明,随机森林模型的精度更高,适用性更强。以30 min为间隔的站点租还车需求预测精度最高,考虑站点土地利用类型后能有效提高模型的预测精度。本文结果可作为未来站点平衡调度的依据并推广应用于共享单车系统,为改善服务水平提供技术和理论支撑。展开更多
基金supported by the National Natural Science Foundation of China(Grants No.41971359)Thanks to the Chengdu Traffic Management Bureau for providing data support for this article.
文摘The successful application of new technologies such as remotely piloted aircraft systems,distributed electric propulsion systems,and automatic control systems on electric vertical take-off and landing(eVTOL)aircraft has prompted Urban Air Mobility(UAM)to be mentioned frequently.UAM is a newly raised transport mode of using eVTOL aircraft to transport people and cargo in urban areas,which is thought to share some of the traffic on the ground.One of the prerequisites for UAM to operate on a regular basis is that its demand can support the operating costs,so forecasting UAM demand is necessary.We conduct UAM demand forecasting based on the four-step method,focusing on improving the third-step modal split,and propose a demand forecasting model based on the logit model.The model combines a nested logit(NL)model with a multinomial logit(MNL)model to solve the problem of non-existent UAM sharing rates.We use Chengdu,China as an example,and focus on forecasting the UAM traffic demand in 2030 with the help of the four-step method.The results show that UAM is suitable for shared operation during the early stages.With a fully shared operation,the UAM share rate increases by 0.73%for every kilometer increase in distance.Moreover,UAM is more competitive than other modes for delivery distances exceeding 15 km.Finally,using the distributions of the share rate and traffic flow pattern from the simulation,we propose the routes that can be prioritized for UAM operations in Chengdu.
文摘针对高密度固定站点式共享自行车系统启停点分布复杂、区域内停放量变化难以监控、区域间流动特征复杂的问题,提出一种聚类算法和时序预测模型组合的需求预测模型。首先,使用基于平衡迭代降维的层次聚类算法(Balanced Iterative Reducing and Clustering using Hierarchies,BIRCH)对共享自行车的启停点进行聚类分析,完成虚拟站点构造和区域划分;其次,对虚拟站点的行程数据进行集计,获得的站点净流入量和站点间流量序列作为输入,使用三次指数平滑法(Triple Exponential Smoothing,TES)进行需求预测;最后,选取纽约和旧金山湾区数据集进行对比和验证。对比结果表明,需求预测模型可有效减少预测单位数量,并准确预测共享自行车在不同区域的供需平衡状态和区域间流动状态。验证结果表明,在2种数据集上,BIRCH算法的聚类质量和耗时均优于K均值算法、层次聚类(离差平方和最小化原则)算法、基于密度带噪声应用的空间聚类和高斯混合模型算法;使用TES模型时预测误差基本小于历史平均模型和自回归移动平均模型。
