【作者】 杨丹；

【导师】 王健；

【作者基本信息】 哈尔滨工业大学 ， 交通运输工程（专业学位）， 2022， 硕士

【摘要】 大型活动的举办严重影响交通系统的正常运行,尤其是在散场后观众同时试图从活动场馆的出口离开时。部分观众没有私家车,产生大量出租车和网约车出行,这是严重交通拥堵的主要原因之一。此外,场馆周围的交通拥堵和车辆不足往往导致乘客等待时间过长。本文在上述背景下,明确大型活动定义、分类和散场特性,处理大型活动当天真实的出租车、网约车数据,从宏观和微观两个层面对大型活动散场的交通影响进行分析,并设计了一种新型的定制公交(customized bus,简称CB)服务,用于服务大型活动的散场观众,并对该定制公交服务全流程进行详细地介绍。针对设计的定制公交服务,本文提出了一种集成数据驱动和需求响应的两阶段规划方法来予以实现。第一阶段将通过数据驱动的方式生成备选定制公交站点。首先提出了一种基于图半监督学习的潜在需求识别模型,通过挖掘观众到达阶段的出租车和网约车出行数据来获取潜在的定制公交需求。然后,基于识别结果,构造CSDBSCAN算法生成聚集式定制公交站点以覆盖聚集需求,并进一步提出CLSDD算法生成离散式定制公交站点,该算法可以与聚集式站点形成拼车路径以覆盖离散需求。第二阶段是需求响应阶段。可供选择的定制公交站点以及相应的票价和承诺时间将提供给活动观众供其选择。分别构建了基于最短路径的单车型线路设计模型和多车型线路设计模型,这两种模型均在用户时间窗和车辆容量约束下,根据最终订单生成具有最大运营商利润的定制公交线路。最后对2019年11月23日于中国天津举行的陈粒演唱会进行案例研究,以评估本文提出的方法。首先,对潜在需求识别结果和生成的定制公交站点进行了详细的分析和结果验证,结果表明识别模型具有较高的准确性,定制公交站点部署合理。然后,为了客观地评估生成的路线和服务,分别设计了不同的定制公交观众选择乘坐比例场景及采用随机抽样的方式生成不同选择场景下的具体乘坐需求,共进行15组不同订单需求下的评估实验,求解多车型线路设计模型生成最终定制公交线路。结果表明,在车内时间差异不大的情况下,定制公交服务比出租车、网约车平均节省总出行时间的37.71%,为每位乘客节省出行费用的36.56%,节省空间道路占用率的89.58%,可以极大地缓解交通拥堵。最后对比分析了单车型线路设计模型和多车型线路设计模型的求解效果,结果表明多车型线路设计模型可以为运营商带来更多的利润。本文的研究能在大型活动散场后为交通管理者提供更科学的定制公交、出租车和网约车的调度依据,既可以解决观众乘车难的问题,又可以有效地缓解交通拥堵;同时为定制公交站点布设和线路设计提供更合理的方法。更多还原

【Abstract】 Planned special events,such as concerts and sporting events,will seriously affect the normal operations of the transportation system,especially at the egress stage when the attendees simultaneously attempt to pour out of the venue.The departing attendees without private cars leads to huge number of taxi and ride-hailing trips,which is one of the prime reasons for serious traffic congestion.Besides,the traffic congestion and insufficient vehicles around the venue often lead to excessive passenger waiting time.Under the above background,this paper clarifies the definition,classification and departure characteristics of planned special events firstly,processes the real taxi and ride-hailing data on the day of planned special events,analyzes the traffic impact of planned special events from the macro and micro levels,and designs a novel customized bus(CB)service for departing attendees of planned special events,then introduces the whole process of the CB service in detail.This paper designs a two-phase planning methodology integrating data-driven and demand-responsive is proposed to implement such a service.The first phase will generate the alternative CB stops through the way of data-driven.An identification model based on graph semi-supervised learning is proposed first to obtain the potential CB demand by mining taxi and ride-hailing trips at the stage of attendee arrival.Then,based on the identification results,a CSDBSCAN algorithm is constructed to generate cluster-type CB stops to cover the clustered demands,and a CLSDD algorithm is further proposed to generate discrete-type CB stops,which can form a carpool path with existing cluster-type stops to cover discrete demands.The second phase is demand-responsive.The alternative CB stops,together with fares and promised time,will be provided to attendees for their choice.The single-vehicle route design model and the multi-vehicle route design model based on the shortest path are proposed respectively,which can generate routes with maximized operator profits under the constraints of user time window and vehicle capacity based on the final orders.A concert held in Tianjin,China is selected to conduct the case study,to evaluate the proposed methodology and the designed CB services.Firstly,the potential demand identification results and the generated CB stops are analyzed and verified in detail,the results show good identification accuracy and reasonable CB stop deployment.Then,in order to objectively evaluate the generated routes and services,15 sets of orders generated by random sampling under different choice ratios,assuming that different proportions of viewers will choose to ride the CB.And CB routes are generated using a multi-vehicle route design model.The evaluation results show that in the case of little difference in in-vehicle time,the generated CB services can save 37.71% of the average total travel time compared with the corresponding taxis and ride-hailing trips,36.56% of the travel expenses for each passenger,and89.58% of the spatial road occupancy,which can greatly alleviate traffic congestion.Finally,the solution effects of single-vehicle route design model and multi-vehicle route design model are compared and analyzed.The results show that the multi vehicle route design model has certain advantages.The research in this paper can provide traffic managers with a more scientific basis for the scheduling of customized buses,taxis and ride-hailing after large-scale events.It can not only solve the problem that it is difficult for spectators to ride,but also effectively alleviate traffic congestion.At the same time,it provides a more reasonable method for customized bus stop layout and route design更多还原