金辉
入学时间：2015级
答辩时间：2019年
论文题目：公共汽车停靠站协同优化方法
中文摘要
摘要
公共交通以其高承载率成为缓解城市交通供需矛盾的重要出行方式，其中公共汽车具有服务成本低、线路及站点设置灵活、建设周期短等特征，适用于绝大多数城市，但现实中公共汽车交通正普遍面临乘客流失的挑战。本研究面向乘客的出行需求，从公共汽车服务的可达性与高效性两个层面，基于公共汽车服务的基础——停靠站，以协同其形式与功能上的核心元素即乘客需求、道路条件和混合车流为基础框架，集成地优化停靠站布设，为从根本上提升公共汽车的服务品质提供理论基础。
主要研究内容和创新点如下：
(1)综合构建了轨道与公共汽车服务半径的影响因素模型。采用低密度条件下、数据完整的家庭出行调查数据，分析了不同公共交通模式、不同区域类型下影响乘客接受较大服务半径的因素；通过统计描述、K值聚类和二元Logistic回归，量化了各种因素对乘客接受较大服务半径的影响；确定了乘客对较大公共交通服务半径的接受意愿，并比较了轨道与公共汽车停靠站半径影响因素的差异；给出了扩大公共交通服务半径、推进建设协作式多模式公共交通系统的政策建议。
(2)提出了路段-节点式、双车道元胞传输模型（Cell transmission model, CTM）。针对驻站公共汽车妨碍或者引起的社会车辆强制换道和自主换道行为，将元胞内的交通流进行成分划分，精细刻画停靠站周边混合车流的动态，并利用流量守恒定律，更新元胞内的车辆数量。利用实际公共汽车的地理位置系统数据、视频监测数据等信息进行模型参数标定，计算结果表明了该模型的结果与实际车流、换道次数的分布趋势吻合，验证了模型的有效性与准确性。此外，对上述模型展开敏感性分析，讨论了不同交通流饱和度下，各种公共汽车停靠站布设方案对混合车流动态的影响，发现存在最小化混合车流延误的停靠站布设方案。
(3)通过一系列二元变量和分段函数，将上述路段-节点式、双车道CTM模型转化为混合整数线性规划模型，采用分支界定法进行模型求解。利用实验线路验证了模型的有效性，给出了一套在线路层面上集成了交通流动态的停靠站布设优化方法，通过VISSIM仿真验证，发现该方法可在多种情况下减少混合车流延误，尤其在路段较短、停靠站密集和交通流饱和度高的情境下有显著效果。
(4)针对同一进口道上多流向公共汽车线路，提出了基于混合交通流动态的停靠站位置与车道功能协同优化模型。采用中微观水平的、基于车道的交通流模型反映停靠站周边车流的动态，通过二元变量和分段函数将模型线性化，建立混合整数线性规划模型。实例表明：该模型可有效降低公共汽车和社会车流的延误；应充分考虑社会车流需求、公共汽车服务频率、驻站时间，以及公共汽车满载率、换乘需求等因素，综合调整停靠站位置，并谨慎采用公共汽车专用道，以实现公共汽车流与社会车流的双赢。
综上所述，本研究从宏、中、微观多个层面为停靠站的规划与设计提供集成性理论基础。
关键词：公共汽车停靠站，乘客需求，可达性，高效性，混合交通流
英文摘要
ABSTRACT
Public transit, characterized with high occupancy, isa well-accepted remedy toalleviate theconflict between travel demand and supply. Compared to rail service, bus service with cost efficiency, design flexibility at both route and stop level, and short contruction period, acts as the backbone of public transit in most areas. However, bus service is generally struggling to retain ridership. The study is oriented with passenger demand, aiming at the two critical attriutes of transit service quality, i.e. accessibility and efficiency, to enhance bus service attraction. Passenger demand, road conditions and mixed traffic low around bus stops are coordinated to synchronically optimize stop location to lay theoretical foundation for bus service improvement.
Major contents and contributions are summarized as:
(1)Construct a comprehensive model of the factors in rail and bus stop service radius. Employing national household travel survey in low-density area with complete data, factors are analyzed of rail andbusstop service radius in various area types. With statistical description, K-value cluster and binary Logistic regression, factors of stop service radius and their impact on service radius are identified and quantified, respectively. Odds chance of passengers accepting large service radius is calculated and the difference between rail and bus service is compared. Suggestions are provided for respondent agencies to enlarge service radius and to promote multimode transit system.
(2)Link-node two-lane based cell transmission model is proposed. As buses dwell at stops, roads along the target bus route is divided into homogeneous sections with bus stop length as the unit, so as to trace the dynamics of buses and cars with mandatory or discretionary lane changing. Based on traffic conservation rule, cell occupation is updated. Data is employed to calibrate the model of bus automatic vehicle location, traffic vedio and three-dimension map, after which the model is validated to precisely capture dynamic traffic flow and lane changing. Sensitity analysis follows to discuss the impact of bus stop location on traffic dynamics under various traffic saturation to find that mixed traffic delay can be minimized with adjusted stop location.
(3)The above link-node two-lane based CTM is approximated with a series of binary variables and piecewise function to convert it into mixed-integer linear programming to explore for the optimal bus stop layout, which is readily solved with standard branch-and-bound method. Vissim simulation for the experimental bus corridor shows that the method can efficiently reduce mixed traffic delay especially with short links, densely spaced stops and highly saturated traffic flow.
(4)Targeting at multi-directed bus stops on the same intersection approach, an optimization model is proposed for bus stop locations embedded with mixed traffic dynamics and lane assignment. Mixed-integer linear programming is established based on the meso-microscopic lane-based traffic flow, with a set of binary variables and piece-wise function. Empirical research shows that the model can effectively reduce bus and car delay. Suitability of exclusive bus lane is comprehensively decided with car demand, bus frequency and dwell time as well as bus loading factor and passenger transfer demand, to work for win-win situation of bus and car flow.
In summary, this dissertation strategically and tactically provides integrated theoretical foundation for stop planning and design.
Key Words:bus stop, passenger demand, accessibility, efficiency, mixed traffic flow
王一喆
入学时间：2015级
答辩时间：2020年
论文题目：智能网联车辆环境下多交叉口协同控制机制研究
中文摘要
摘要
常规车辆、与交通控制系统无交互功能的联网车辆以及与交通控制系统有交互功能的智能网联车辆组成的新型混合交通流已呈不断发展趋势并将持续存在，城市道路网络交通流结构亦随之发生重大改变。本研究思路正是源于此背景下的迫切需求而提出，围绕联网车辆环境下多交叉口协同控制机制、智能网联车辆环境下交叉口主动控制机制、智能网联车辆环境下多交叉口协同控制系统及实验平台三大主线内容展开，面向智能网联车辆间断型混合交通流，通过解析各类车辆跟驰、换道等驾驶模式，研究不同混合条件下智能网联车辆混合交通流特征及信息提取方法，以最小协调控制网络——多交叉口为研究对象，整合交通信号灯控制、智能网联车辆主动控制及通过智能车辆间接控制常规车辆三种控制模式，研究智能网联车辆混合交通流环境下主动控制机制及多交叉口协同控制策略。结合真实车辆及交通控制系统验证研究协同控制理论，构建智能网联车辆交通流环境下信息交互及协同控制集成实验平台。
联网车辆环境下多交叉口协同控制机制。借助联网车辆精准轨迹数据等非集计型数据和交警线圈卡口等集计型数据进行多源数据融合，分别从周期、相位差和绿信比三个交通信号控制关键参数入手，构筑基于人工智能与交通工程专业相结合的分层递阶控制算法。一是通过交叉口间车辆起止点分布等特征参数对控制子区进行划分，并综合考虑不同信号控制效果的评价指标，借助基于模型的算法获得最佳周期；二是利用基于高斯过程的贝叶斯最优化方法，优化多交叉口干线协调相位差，进而从交通工程的专业视角设置不同的绿波带宽以匹配上下行流量、运行速度不均匀的潮汐现象；三是利用Q学习等智能算法借助流量、密度、速度等关键交通流参数，对多交叉口绿信比进行迭代优化。交通子区划分，周期优化，相位差优化和绿信比优化，每一层之间都是强关联，互相进行信息传递和效果反馈，进而形成基于分层递阶的多交叉口协同控制算法，使城市交通系统的可测性与可控性逐渐提升。相较于多时段定时控制和既有自适应控制，可显著提升包括多交叉口总延误、总行程时间、总停车次数、驶离路网车辆数在内的全局效益。进一步分析得出，基于分层递阶的交通信号控制算法适用于高流量和多交叉口等复杂的交通环境，但并不适用于低流量和单交叉口等简单的交通环境，因此在解决交通问题时不能不加区别地应用人工智能型算法。
智能网联车辆环境下交叉口主动控制机制。联网车辆的特点是可以提供精准的轨迹级交通信息，局限是只能向交通控制系统单方向提供信息，无法接收来自于交通控制系统的反馈。本研究通过构建智能网联车辆环境下交叉口主动控制机制，探索解决该问题的方法。一是在联网车辆环境下多交叉口协同控制机制的基础上，得出多交叉口信号配时最优方案，借助人工智能、大数据、云计算等前瞻技术，沿用数据、通信、算力、算法相结合的思路，助力“聪明的车”，保障驾驶员和车内乘客的出行效率与安全。二是借助智能网联车辆与交通控制系统的信息交互能力，实现被控制方适应控制方——智能车速引导和控制方适应被控制方——智能网联车辆信号优先，为驾驶员提供其肉眼观察和经验判断所无法获取的信息，打造“智能的路”。其中：智能车速引导功能包括效率型、经济型和舒适型车速引导策略，可以根据不同的需求为客户提供行驶速度的范围；智能网联车辆环境下交叉口优先控制方案包括绿灯相位延长、红灯相位缩短和绿灯相位插入三种措施，可根据智能网联车辆到达交叉口的时刻采取不同的优先方案,同时设置智能网联车辆信号优先判定方法，从系统最优的角度决定是否实施智能网联车辆信号优先，防止出现“浪费式”优先。三是提出通过智能网联车辆间接控制和影响常规车辆的方案，对比不同智能网联车辆渗透率和相同渗透率不同车队模式条件下智能网联车辆对常规车辆的影响，研究揭示了智能网联车辆最佳渗透率为30%-40%，可实现通过少数智能网联车辆充分影响和间接控制更多常规车辆的边际效益最优化。进一步针对交通信号灯控制、智能网联车辆主动控制及通过智能车辆间接控制常规车辆等控制模式，以交通信号控制生成系统最优方案为基础，引入智能网联车辆引导及主动控制兼顾用户最优，并将智能网联车辆间接控制常规车辆作为解决不同渗透率和车队模式条件下的过渡状态，实现三种控制模式互相协同。
智能网联车辆环境下多交叉口协同控制系统及实验研究。本研究通过对联网车辆环境下多交叉口协同控制机制以及智能网联车辆环境下主动控制机制进行实地验证，结合高精度地图资源打造智能网联车辆环境下多交叉口协同控制系统及实验平台，并将原创性的算法应用在广州市中心城区实际交通控制系统中。其创新点和核心优势为：一是构筑了可复用性强、自动化程度高的间断交通流信息提取与演化规律分析方法以及层次清晰、易于输入输出和查询的存储结构，提高数据可信度及数据缺失的容错度；二是通过交通子区的划分避免求解过程陷入维度灾难，通过设置合理的奖惩函数及闭环反馈机制，促进区域级交通信号控制更好的产业化推广；三是设计自动化程度较高的分层递阶式智能交通控制机制，将智能算法与交通工程专业知识相结合，从宏观、中观、微观三个维度打造优化系统，进而实现“时间上争分夺秒，空间上寸土必争”的区域级多交叉口协同优化平台。本平台的前瞻性在于：面向未来10-20年最常见的常规人类驾驶车辆、联网车辆与智能网联车辆混行场景，提供在保障有限用户最优条件下实现系统最优的协同管控方案，为未来的全自动驾驶交通场景打下基础。
随着智能网联车辆和车路协同技术的进一步发展和推广应用，人车路一体化的交通系统将成为现实，在全时空交通信息的协同感知、融合和交互的条件下，研究智能网联车辆交通流环境下多交叉口协同控制理论、方法及其关键技术已成为必然。智能网联车辆环境下交通控制机制的最终目标是打造“聪明的车+智能的路”，利用智能网联车辆等技术将“车”和“路”有机的结合在一起，解决交通系统的“顽疾”，为城市打通“任督二脉”。综上所述，本研究具有清晰的方向性、前沿性及趋势性，对进一步推动新一代交通网络控制系统的发展和应用，为我国抢占该领域的制高点具有重要的战略意义和实际价值。
关键词：智能网联车辆，多交叉口，分层递阶式控制，交通主动控制机制，协同控制实验，车路协同，智能交通系统
英文摘要
ABSTRACT
A new heterogeneous traffic flow consists of regular vehicles, internet vehicles having no interactive functions with traffic control systems, and intelligent-connected vehicles having interactive functions is updating the composition of the current urban road network traffic flow. It has been a growing trend and will continue to be so.On account of the urgent demand, the research focused on three main parts of multi-intersection coordinated control mechanisms in the internet vehicle environment, intersection active control mechanism in the intelligent-connected vehicle environment, and multi-intersection coordinated control system and experimental platform in the intelligent-connected vehicle environment.1) For heterogeneous interrupted traffic flow of intelligent-connected vehicles, to analyze the characteristics and information extraction method of heterogeneous traffic flow of intelligent-connected vehicles under different conditions, the research examined driving modes of three types of vehicles, including car following and lane changing. 2) This study integrated three control modes of traffic signal control, active control of intelligent-connected vehicles, and indirect control of regular vehicles through intelligent vehicles to study the active control mechanism and multi-intersection coordinated control strategy for intelligent-connected vehicle heterogeneous traffic flow. The control object is a minimum coordinated control network – a multi-intersection. 3) With the combination of real vehicles and traffic control systems, this work validated coordinated control theory and built an integrated experiment platform of information interaction and coordinated control under intelligent-connected vehicle heterogeneous traffic flow environments.
The coordinated control mechanism of multiple intersections in Internet vehicles environment.By multi-source data fusion with non-aggregated data (e.g., precise trajectory data of internet vehicles) and aggregated data (e.g., traffic police coil chokepoints), a hierarchical control algorithm based on the combination of artificial intelligence and traffic engineering is constructed, focusing on the three key parameters of the cycle, arterial coordination signal offset and green split.First of all, according to the characteristic parameters (e.g., OD distribution of vehicles between intersections), the sub-regions are divided. Then, integrate indexes for evaluating the effectiveness of different signal control and obtain the optimal period with the help of model-based algorithms.What's more, using the Bayesian optimization method based on the Gaussian process, optimize the phase offset of the arterial coordination control, and set different green wave bandwidths to match the upstream and downstream traffic and the tidal phenomenon with uneven travel speed from the professional perspective on traffic engineering.Finally, the intelligent algorithm, such as Q learning, is used to iteratively optimize the green split of multi-intersection by using key traffic flow parameters such as traffic flow, density, and speed.There is a strong association between the sub-region division, the period optimization, the phase offset optimization, and the green split optimization. With information transfer and effect feedback, the hierarchical control algorithm for multi-intersection coordinated control is formed, which gradually improves the measurability and controllability of the urban traffic system.Compared to multiple intervals fixed-time control and existing adaptive control, the hierarchical control algorithm can significantly improve multi-intersections overall benefits, including total delay, total travel time, the total number of stops, and the number of vehicles leaving the road network.Furthermore, the hierarchical control algorithm for multi-intersection coordinated control is suitable for complex traffic environment with high volumes and multiple intersections but not ideal for a simple traffic environment with low volumes and one single intersection. Therefore, artificial intelligence algorithms cannot be applied indiscriminately in traffic problems.
The active control mechanism of intersection in intelligent-connected vehicles environment.Internet vehicles are able to provide accurate track-level traffic information, with the limitation to one-way information provision to the traffic control system and no feedback reception from the traffic control system.This study explored the solution to this problem by constructing an active control mechanism for intersections under an intelligent-connected vehicle environment.Firstly, based on multi-intersection coordinated control mechanism in an internet vehicle environment, the optimal multi-intersection signal timing program is conducted, with the help of some forward-looking technologies (e.g., artificial intelligence, large data, cloud computing) and the combination of data, communication, computing power, and algorithms, which will be to protect the efficiency and safety of the driver and passengers in the "smart vehicle".Secondly, using the information interaction between intelligent-connected vehicles and traffic control system, above the "intelligent infrastructure", achieve both the adaption of the controlled side to the control side - intelligent speed guidance, and the adaption of the control side to the controlled side - intelligent-connected vehicle signal priority, and provide drivers the information beyond their naked eyes and experiences. The intelligent speed guidance function includes efficient, economical, and comfortable speed guidance strategies, which can provide customers with a range of conversion speeds according to different demands; The priority control scheme of intersections in the intelligent-connected vehicle environment includes green light phase extension, red light Phase shift, and green light phase shift measures can take different priority schemes according to the time when the intelligent-connected vehicle arrives at the intersection. Simultaneously, the intelligent-connected vehicle signal priority decision method is set. From the perspective of system optimization, decide whether to implement intelligent-connected vehicle signal priority to prevent "wasteful" priority.Thirdly, it is proposed to control indirectly regular vehicles through intelligent vehicles. Comparing the influence of intelligent-connected vehicles on regular vehicles under different intelligent-connected vehicle permeability and different vehicle fleet modes with the same permeability, the research reveals that at 30%-40% permeability the indirect control of regular vehicles through intelligent-connected vehicles gets the optimal marginal benefit, which means that make a full influence on and control indirectly as more regular vehicles as possible through as less intelligent-connected vehicles.Then, focusing on the control modes of traffic signal control, active control of intelligent-connected vehicles and indirect control of regular vehicles through intelligent vehicles, the study balances system optimization and user optimization, based on the system optimum solution derived from traffic signal control and on guidance and active control of intelligent-connected vehicles. Finally, indirect control of regular vehicles through intelligent vehicles is used as a solution to the transition under different permeability and vehicle fleet modes, achieving coordinated control among the three control modes.
Multi-Intersection coordinated control system and experimental research in Intelligent-connected vehicle environment.In this study, the multi-intersection coordinated control mechanism in the internet vehicle environment and the intersection active control mechanism in intelligent-connected vehicle environment are put into practice. The multi-intersection coordinated control system and experimental platform in an intelligent-connected vehicle environment are built with high-precision maps. The original algorithm is applied to the actual traffic control system in downtown Guangzhou.The following are the innovations and core advantages: first of all, a highly reusable and automated method to extracting information and analyzing evolutionary patterns for discrete traffic flow and a clearly hierarchical storage structure easy to input, output and query are built to improve data credibility and fault tolerance for missing data;Second of all, dimensional disaster in the process of solving can be avoided through the division of sub-regions, and Industrialization promotion on traffic signal control at the regional level can be better through the reasonable reward and punishment functions and closed-loop feedback mechanism;The third one is combining intelligent algorithms with traffic engineering, a hierarchical control mechanism for intelligent traffic control with a higher degree of automation has been designed to create an optimization system from three dimensions - macro, meso, and micro. Then, a regional multi-intersection coordinated optimization platform is achieved towards full and efficient utilization of space and time resources.Which the foresight of this platform is that it can provide a coordinated control program to realize the optimal system under optimal conditions for limited users, facing the most common scenario that the heterogeneous traffic flow consisting of regular human-driven vehicles, internet vehicles and intelligent-connected in the next 10-20 years, and lay the foundation for future fully automated driving traffic scenarios.
With further development and promoting application of the intelligent-connected vehicles and cooperative vehicle infrastructure technologies, an integrated human-vehicle-infrastructure transportation system will come true.In the condition of cooperative perception, fusion, and interaction of all space-time traffic information, it is urgent to research the theories, methods, and crucial technologies for coordinated control of multi-intersections under the environment of Intelligent-connected vehicle traffic flow.The ultimate goal of the traffic control mechanism in the intelligent-connected vehicle environment is to create an intelligent transportation system, using technology such as intelligent-connected vehicles and achieving the organic combination between "smart vehicle" and "intelligent infrastructure", to ensure the proper functioning of the urban transportation system.As mentioned above, this study has a clear direction, frontier, and trend, which is of strategic significance and practical value to further promote the development and application of the next-generation traffic network control system and for China to seize the high ground in this field.
Key Words:Intelligent-Connected Vehicle，Multi-intersection，Hierarchical control algorithm，Active control mechanism，Coordinated control experimental platform，Cooperative Vehicle Infrastructure System，Intelligent transportation system
