赖金涛
入学时间：2019级
答辩时间：2023年
论文题目：基于轨迹的多交叉口新型混合交通流协同控制理论研究
中文摘要
摘要
车路协同作为我国多个科技战略的聚焦点，推动了城市道路交通的网联化和智能化发展，使传统交通环境发生了重大变革并涌现出由人类驾驶车辆（Human-driven Vehicle, HV）和网联自动驾驶车辆（Connected and Automated Vehicle, CAV）共同构成的新型混合交通流。新型混合交通流具有泛随机、强交互、非均质等特征，会减小道路通行能力且降低交通需求的可控性，进一步加剧了城市道路交通供需错配的问题。传统基于集计交通流的交通控制理论与技术，已经难以适用于新型混合交通流的控制，故亟需变革现有交通控制理论。针对这一主要矛盾，本文面向城市道路多交叉口场景，充分利用新型混合交通流中车辆轨迹的可知性和可控性，以多交叉口连续道路空间上供需协同与平衡为目标，以“改变车辆轨迹，调节交通需求，优化供给分配”为核心学术思想，研究基于轨迹的多交叉口新型混合交通流协同控制理论。研究依照“交通机理解析—需求调节机制—供给分配机制—供需协同理论”的思路展开。
在交通机理解析层面，针对现有新型混合交通流机理解析缺少实证支撑、交通流实验环境不可信的问题，构建一套可信的新型混合交通流环境分析框架以及一套基于轨迹数据的虚拟环境标定方法。该环境分析框架面向交叉口大规模新型混合交通流场景，基于博弈理论和优化控制方法，解析网联自动驾驶车辆与人类驾驶车辆高动态交互的关系，设计车辆轨迹对于城市道路交叉口控制的响应模型，构建出逼真的新型混合交通流虚拟环境。针对该虚拟环境，考虑环境参数降维及环境加速标定的需要，构建基于强化学习的三阶段环境参数自整定算法，并利用实车采集的轨迹级数据，对虚拟环境参数进行标定和校准。实验表明，该环境分析框架及标定方法能够高精度模拟宏观交通流和微观车辆行为，保证新型混合交通流虚拟环境的可信度及可用性。
在需求调节机制层面，针对现有需求控制方法过度强调交通机动性，而忽略轨迹调控对交通流造成负面影响的问题，建立一套微观轨迹-宏观车流协同的交叉口瓶颈主动需求调节方法。该主动需求调控方法能够在保障微观车辆机动性的同时降低对交通流的负面效应。它以“车流-轨迹双向协同”为核心思想，提出多中心式的主动需求调节框架，将交叉口瓶颈控制问题被分解为多个道路片段的需求调节问题，通过协调多个路侧中心的控制效能，保障多个道路片段上车辆轨迹的连续性以及交通流运行的连贯性。在此基础上，构建迹流协同的需求调节模型，以交通流需求主动追踪交叉口瓶颈通行能力为核心目标，通过优化调节微观车辆轨迹，得到最佳的交通流需求到达模式，从而降低交叉口瓶颈控制对交通流的负面效应。实验表明，相较于传统需求调控方法，本研究的控制方法能够有效提升交叉口瓶颈的需求控制精度，并显著减小车辆巡航不适度以及交通流扰动等控制负面效应。
在供给分配机制层面，针对传统交叉口供给分配方法在道路时空资源分配上不精准、不灵活的问题，建立轨迹-路权协同的交叉口自适应供给分配方法。该方法将交叉口连续道路上的车辆间隙看作可供分配的时空路权，基于轨迹信息辨识未被使用的道路时空路权，并面向不同优先级的交通需求，设计灵活的局部路权分配机制。在此机制下，构建轨迹-路权协同决策模型，通过局部轨迹调整与路权分配动态之间的协同，将局部路权精准分配给关键授权主体，实现交叉口道路供给资源的动态高效分配。实验表明，相比于现有方法，本研究方法在提高交叉口交通通过量、降低车辆平均油耗以及降低车辆平均延误等方面均具有明显优势。
在供需协同理论层面，面向新型混合交通流加剧多交叉口供需错配的现状，针对传统断面式控制方法难以遏制连续道路空间上供需矛盾的难题，建立基于轨迹的多交叉口交通供需协同控制理论框架。该理论框架将多交叉口供需协同问题拆解为三类子问题，并建立这三类问题的联动机制：1）交叉口断面供给分配。考虑了上下游交叉口协调关系，设计基于双环屏障控制的自适应信号控制方法，利用双层动态规划对信号控制方案进行优化，最小化交通延误时间。2）连续道路供给再分配。考虑道路资源的时空占用情况，基于过饱和度识别高需求转向车流，引入上述轨迹-路权协同的自适应供给分配机制，允许高需求转向车流使用多功能车道的通行路权，以疏散高需求车流。3）基于轨迹的主动需求调节。考虑交叉口断面供给以及连续道路路权的约束，建立车辆轨迹平滑模型，引入上述车流-轨迹协调的主动需求调节机制，实现交通需求主动匹配断面通行权和局部路权。最终，在上述新型混合交通流虚拟环境中，进行多交叉口供需协同控制实验的验证。实验结果表明，与传统断面式控制方法相比，本研究提出的控制方法，在通行效率、安全性以及可持续性等层面具有显著优势。
本文从多交叉口供需错配矛盾出发，基于交通机理解析，构建了新型混合交通流虚拟环境与场景，分别从需求和供给层面提出了车流-轨迹协调的主动需求调节机制以及轨迹-路权协同的自适应供给分配机制，构建了基于轨迹的多交叉口供需协同控制理论框架，为未来新型混合交通流控制技术的落地应用奠定了理论基础。
关键词：新型混合交通流，轨迹级交通控制，主动需求调节，自适应供给分配，交叉口供需协同
英文摘要
ABSTRACT
As a focal point of technological strategies in China, Cooperative Vehicle Infrastructure System (CVIS) has promoted the connected and intelligent development of urban road traffic. It leads to significant changes in the traditional transportation environment and the emergence of partially connected and automated traffic composed of Human-driven Vehicles (HV) and Connected and Automated Vehicles (CAV). The partially connected and automated traffic exhibits characteristics such as randomness, strong interaction, and heterogeneity. Such characteristics cause the reduction of road capacity and traffic demand controllability, and further exacerbate the mismatch between supply and demand in urban road traffic. Traditional traffic control theories based on aggregated traffic flow are no longer applicable to the control of partially connected and automated traffic. Thus, the existing theories and technologies of traffic control urgently need to be reformed. Due to the reformation need, this study focuses on urban road intersection scenarios and fully utilizes the predictability and controllability of vehicle trajectories in partially connected and automated traffic. The study aims to achieve supply-demand coordination and balance on continuous road spaces with multiple intersections. The research idea is "changing vehicle trajectories, regulating traffic demand, and optimizing supply allocation". This study establishes trajectory-based cooperative control theory for the partially connected and automated traffic at multiple intersections. The study follows the train of thought of "traffic mechanism analysis - demand regulation mechanism - supply allocation mechanism - supply-demand coordination theory."
From the aspect of traffic mechanism analysis, the lack of real-world validation for partially connected and automated traffic mechanisms and the incredibility of experimental environments should be addressed. Thus, a credible analysis framework for a partially connected and automated traffic environment and a trajectory-based environment calibration method are constructed. The environment analysis framework is designed for large-scale partially connected and automated traffic scenarios at intersections and utilizes game theory and optimization control methods to analyze the intensive interaction between CAV and HV and model the responses of vehicle trajectories to intersection control. This framework constructs a realistic virtual environment for partially connected and automated traffic. To meet the needs of environment parameter dimensionality reduction and accelerated calibration, a trajectory-level environment parameter self-tuning algorithm based on reinforcement learning is developed. Trajectory data collected from real vehicles are used to calibrate and validate the virtual environment parameters. Experimental results demonstrate that the proposed framework and method can accurately simulate macroscopic traffic flow and microscopic vehicle behaviors, ensuring the credibility and availability of the virtual environment for partially connected and automated traffic.
From the aspect of the demand regulation mechanism, the problem that existing demand control methods overemphasize traffic mobility and neglect the negative impacts on traffic flow caused by trajectory adjustments should be addressed. Thus, an active demand regulation method that coordinates micro-trajectories and macro-flow at intersections is established. This method reduces the negative effects on traffic flow while ensuring individual mobility. With the core idea of "trajectory-flow coordination," a multi-center demand regulation framework is proposed to decompose the intersection bottleneck control problem into several sub-problems for multiple road segments. By coordinating the control efficiency of multiple roadside centers, the continuity of vehicle trajectories and traffic flow on multiple road segments is ensured. Based on this, a trajectory-coordinated demand regulation model is constructed to actively regulate traffic flow demand to track the intersection bottleneck capacity. By optimizing the trajectories of vehicles, the optimal traffic flow demand arrival pattern is obtained and the negative impacts of intersection bottleneck control on traffic flow are reduced. Experimental results show that compared to traditional demand control methods, the proposed control method in this study can effectively improve the precision of intersection bottleneck control and significantly reduce negative traffic impacts in terms of vehicle cruising discomfort and traffic flow fluctuation.
From the aspect of the supply allocation mechanism, the inaccuracy and inflexibility of traditional intersection supply allocation methods should be addressed. An adaptive supply allocation method with trajectory-and-right-of-way coordination is established. This method considers the vehicle gaps on the continuous road space of intersections as spatiotemporal road rights-of-way to be assigned. It identifies unused rights-of-way based on trajectory information and designs a flexible right-of-way allocation mechanism for traffic demands with various priorities. Under this mechanism, a trajectory-and-right-of-way coordinated decision model is constructed. It dynamically and efficiently allocates local rights-of-way to key authorized entities by coordinating local trajectory adjustments and right-of-way allocation. Experimental results demonstrate that compared to existing methods, the proposed method in this study has significant advantages in increasing traffic throughput at intersections, as well as reducing fuel consumption and vehicle delay.
From the aspect of supply-demand coordination theory, it is considered that the significant mismatch between supply and demand is caused by partially connected and automated traffic at multiple intersections. To address the challenge of the supply-demand mismatch in continuous road spaces, which is a puzzle in traditional section-based control methods, a trajectory-based supply-demand coordination control theory framework is established for multiple intersections. This theory framework decomposes the supply-demand coordination problem at multiple intersections into three sub-problems and establishes linkage mechanisms among these sub-problems: 1) Section supply allocation at intersections: Considering the coordination relationship between upstream and downstream intersections, an adaptive signal control method based on double-ring barrier control is designed. Dynamic programming is used to optimize the signal control scheme and minimize traffic delay. 2) Supply redistribution on continuous road space: Considering the characteristics of partially connected and automated traffic, a trajectory-based supply redistribution method is proposed. It dynamically adjusts the allocation of road space resources based on real-time traffic conditions. The method aims to balance the traffic demand and supply on continuous road spaces and improve overall traffic efficiency. 3) Trajectory-based active demand regulation: Taking into account the constraints of intersection section supply and continuous road right-of-way, a vehicle trajectory smoothing model is established. The proactive demand regulation mechanism with flow-trajectory coordination is introduced, achieving proactive matching of traffic demand with section passing rights and local rights-of-way. Finally, in the aforementioned virtual environment, experiments are conducted to verify the multi-intersection supply-demand coordination control theory. The experimental results show that compared to traditional section-based control methods, the control method proposed in this study has significant advantages in terms of traffic efficiency, safety, and sustainability.
This thesis begins with the contradiction of mismatched supply and demand at multiple intersections. It integrates traffic mechanism analysis, demand regulation, supply allocation, and supply-demand coordination. It proposes a theoretical framework for trajectory-based supply-demand coordination control at multiple intersections and lays a theoretical foundation for the future implementation of partially connected and automated traffic control technologies.
Key Words: Partially Connected and Automated Traffic, Trajectory Based Traffic Control, Active Traffic Demand Regulation, Adaptive Supply Allocation, Intersection Supply-Demand Coordination
