【作者】 王鹏飞；

【导师】 陈学武；

【作者基本信息】 东南大学 ， 交通运输工程， 2021， 博士

【摘要】 随着“互联网+”、大数据和云计算等新一代信息技术的发展,公交信息化建设已逐渐从供给信息单向提供向供需信息双向传递过渡。当前,包含车辆位置、预计到站时间、车内拥挤度的供给信息和包含乘客起终点、出发时间的需求信息已能实现实时双向沟通传递。在供需信息实时提供并双向传递的背景下,乘客可以通过供给信息获取更准确的车辆预计行驶时间、更全面的车辆状态(如拥挤度),根据行程安排和个人偏好,合理选择出行方案;运营方可以通过需求信息更科学地指导线路规划、车辆调度,甚至探索开发定制公交、灵活式公交等服务新模式。实时供需信息能让供需双方更精确匹配,为提升城市公交服务水平和乘客出行体验带来了新的发展机遇。本文结合当前技术背景和公交服务创新趋势,以提升实时供需信息条件下城市公交服务水平和乘客出行体验为目标,针对高频公交、低频公交和需求响应式接驳公交,分别从乘客舒适性、车辆守时性和服务易获性等角度,提出供需信息条件下运营优化策略,以期充分利用信息要素提升城市公交运行效率、可靠性、舒适性。本文主要研究工作如下。(1)基于载客量预测的实时公交拥挤度信息发布方法研究。为准确提供实时公交拥挤度信息,首先提出基于自动乘客计数数据的两阶段载客量预测方法,该方法阶段一运用自适应卡尔曼滤波,预测站点层面的(上车、下车和断面)客流,阶段二运用支持向量回归,预测车辆层面的载客量;进而基于预测载客量,提出考虑折减系数的拥挤度信息发布策略,折减系数的存在使得低拥挤度信息发布更加谨慎。以苏州1路公交为案例,测试了所提预测方法和信息发布策略的有效性。结果表明,两阶段载客量预测方法优于现有预测模型,尤其适用于上车主导型区段预测和多步预测,基于载客量预测的拥挤度信息发布策略能够有效提升拥挤度预测能力,考虑折减系数后,拥挤度预测准确率有略微下降,但能获得更高的乘客增益,表明更少出现发布信息与实际不符的状况。(2)基于实时拥挤度信息发布的高频公交运行自控制策略研究。针对高频公交提出一种运行自控制策略,即向乘客发布实时拥挤度信息,简称为拥挤度信息策略。为测试拥挤度信息策略效果,首先建立了拥挤度信息条件下公交运行模型,考虑了乘客需求强度、乘客行为、车辆运行时间等多种随机因素。通过多场景、多需求强度、多策略组合下的仿真实验,详细评估了其对于运行稳定性和乘客出行体验的提升效果。仿真实验表明,拥挤度信息策略能够在多个场景下使公交运行稳定性提升20%,且能使乘客感受的车内拥挤度最多降低25%,发车频率高、路段运行时间不稳定的线路,改善效果更显著;敏感性分析进一步说明,信息渗透率是影响该策略效果的重要因素,但25%的信息渗透率仍能产生理想情形(100%渗透率)约60%的改善效果。基于常州2路公交实际运营数据的案例分析结果与仿真实验相似,验证了该策略应用于实际线路的有效性。(3)低频公交到站时间信息发布和车速控制策略研究。针对低频公交,提出车辆运行与到站信息发布相契合的“相对守时”概念。首先建立双向贝叶斯到站时间预测模型,该模型包含纵向和横向两种贝叶斯算子,其本质是从横纵两个角度寻找与目标班次相似的历史班次,再对到站时间先验概率分布进行修正并综合。根据双向贝叶斯模型所得到站时间分布概率,提出到站信息发布策略,对上车到站信息和下车到站信息予以分开讨论,分别以相应的概率分布系数作为信息发布依据;进一步提出基于减速控制触发阈值的车速控制策略。南京40路公交案例分析表明,双向贝叶斯模型能有效进行到站时间单值预测和区间预测,其中横向贝叶斯算子在车辆接近目标站点时具有显著优势;车速控制策略能够以较小代价提升到站信息准确率或车辆准点率约5%。(4)基于信息交互的需求响应式接驳公交乘客导向运行策略研究。针对含有待选站点的需求响应式接驳公交系统,提出一种基于信息交互的乘客导向运行策略,该策略由预计到站时间更新机制和乘客接纳判别机制所组成,含有两个核心参数:到站时间限制指数和车内时间限制指数,两个参数分别确保所有站点的到站时间均受到乘客上车时间和下车时间的约束,以保证每位乘客的出行质量。为解决该系统中实时站点选取及路径规划问题,首先建立混合整数规划模型,进而针对所研究问题特点,提出一种启发式算法进行求解。算例分析结果表明,相比于传统灵活式公交运营策略和固定线路公交,本文提出的乘客导向策略避免了车辆晚到导致的乘客候车时间,同时有效减少了车内时间,在乘客总费用上具有较明显的优势,优势随着乘客需求强度增大而增大。敏感性分析显示,算例条件下到站时间限制指数和车内时间限制指数的最优取值分别为1.4和2.5。研究成果将丰富信息条件下城市公交运营特征和策略研究,同时可为优化实时信息发布、提升实时供需信息条件下城市公交服务水平和乘客出行体验提供理论基础和方法指导。更多还原

【Abstract】 With the development of the new generation of information technology,such as Internetplus,big data and cloud computing,the bus information service system has gradually moved from providing supply information only to a two-way interaction between supply and demand information.At present,the supply information(e.g.,vehicle location,arrival time and invehicle crowding)and the demand information(e.g.,passenger O-D and departure time)can realize real-time two-way communication.Passengers can obtain more accurate bus arrival time and more comprehensive vehicle status(such as crowding)through the supply information,and reasonably choose the travel plan according to their schedules and personal preferences;Operators can make route planning and vehicle scheduling more scientifically through the demand information,and even explore new service modes such as customized bus and flexible transit.Totally speaking,real-time supply and demand information can make the supply and demand match more accurately,which brings new development opportunities for improving the level of service(LOS)of bus transit and passenger travel experience.Considering current technical background and the innovation trend of bus service,this study aims to improve the LOS and passenger travel experience of bus transit under the condition of real-time supply and demand information.The operational optimization strategies under real-time supply and demand information are proposed from the perspectives of passenger comfort(for high-frequency transit),vehicle punctuality(for low-frequency transit),and service accessibility(for responsive feeder transit).The main research work of this paper is as follows.(1)A release method of real-time crowding information based on bus passenger load prediction.In order to provide accurate real-time crowding information(RCI),a two-stage method for bus passenger load prediction using automatic passenger counting data is proposed.In the first stage,adaptive Kalman filter is used to predict(boarding,alighting and section)passenger flows at station level.In the second stage,support vector regression is used to predict the passenger load at vehicle level.Then,based on the predicted passenger load,a release policy of RCI which contains a reduction factor is proposed.The reduction factor makes the release of low-level crowding information more cautious.Using a case study based on Suzhou No.1bus line,the effectiveness of the two-stage passenger load prediction method and the information release policy is tested.The results show that the two-stage method is superior to the existing prediction models,especially suitable for the boarding dominated segment prediction and multi-step ahead prediction.The RCI release policy based on the predicted passenger load can effectively improve the crowding prediction ability.Considering the reduction factor,the accuracy of crwoding prediction decreases slightly,but it can obtain higher passenger gain,which means there is less inconsistency between the released information and the actual situation.(2)A self-control strategy for high-frequency transit based on provision of real-time crowding information.This study proposes a self-control strategy for high-frequency transit,which is providing real-time crowding information for passengers(noted as RCI strategy).In order to test the effect of RCI strategy,a bus motion model under RCI is established,which considers multiple random factors such as demand intensity,passenger behavior,and vehicle running time on links.Through the simulation experiments of multi scenario,multi demand intensity and multi strategy combination,the effect of RCI strategy on the operation stability and passenger travel experience is evaluated in detail.The simulation results show that RCI strategy can improve the bus operation stability by 20% in each scenario,and reduce the passenger’s feeling of crowdness by 25% at most.The improvements are more significant for routes with high departure frequency and unstable running time on links.Sensitivity analysis further shows that the penetration rate of RCI is an important factor affecting the effect of the strategy,but 25% penetration rate can still produce about 60% improvement compared with the ideal situation(100% penetration rate).The case study based on the actual data from Changzhou No.2 line further verifies the effectiveness of RCI strategy when applied to the actual line.(3)Bus arrival time information release strategy and speed control strategy for lowfrequency transit.The concept of "relative punctuality" is proposed for low-frequency transit.The idea of the concept is to make the vehicle operation and the released information of vehicle arrival time consistent.Firstly,a bi-directional Bayesian model for arrival time prediction is established.The model includes two kinds of Bayesian operators,i.e.,vertical and horizontal.The essence of the model is to find the historical trips that are similar to the target trip from the horizontal and vertical perspectives,and then to modify and synthesize the prior probability distribution of the arrival time.Based on the arrival time distribution probability obtained by the bidirectional Bayesian model,an arrival information release strategy is proposed.The strategy discusses arrival information for boarding passengers and alighting passengers separately,and adopts corresponding probability coefficients as the basis for information release.Furthermore,a speed control strategy based on the trigger threshold of deceleration is proposed.A case study from Nanjing No.40 line shows that the bi-directional Bayesian model can effectively predict the arrival time in terms of single value and interval,and the horizontal Bayesian operator has significant advantages when the vehicle approaches the target station.The speed control strategy can improve the arrival information accuracy rate or vehicle punctuality rate by about 5% at a small cost.(4)A passenger-oriented operation strategy for responsive feeder transit based on information interaction.A passenger oriented operation strategy is proposed for the responsive feeder transit systems with meeting points.The strategy consists of an update mechanism of estimated vehicle arrival time and a discrimination mechanism of passenger acceptance,and contains two key parameters,namely arrival time limit index and in-vehicle time limit index,which ensure that the vehicle arrival times at stops are constrained by boarding time or alighting time of passengers respectively.In order to solve the real-time stop selection and route planning problem in the system,a mixed integer programming model is established,and then a meta heuristic algorithm is proposed to solve the problem.The results show that,compared with the traditional operation strategy of flexible transit and the fixed-route line,the passenger-oriented strategy is able to avoid passengers’ waiting times caused by late arrival,and effectively reduces the in-vehicle times.It has obvious advantages in the total cost of passengers,and the advantages increase with the passenger demand intensity.Sensitivity analysis shows that the optimal values of arrival time limit index and in-vehicle time limit index are 1.4 and 2.5respectively under the numerical experiment.This study will enrich the research regarding bus operation characteristics and strategies under information,and provide theoretical basis and method guidance for optimizing real-time information release and improving LOS and passenger travel experience of bus transit under real-time supply and demand information.更多还原