龚华天

入学时间：2021级

答辩时间：2025年

论文题目：城市道路网络交通生命线优化设计方法研究

中文摘要

摘要

城市道路交通网络作为城市发展的核心组成部分，是确保城市日常运作不可或缺的基础设施。然而，道路网络常因突发事件而受损，严重影响城市应急响应效率及居民生命财产安全。本研究旨在构建系统的道路网络交通生命线优化设计方法，以提升城市在应急事件中的快速响应能力，保障城市安全与可持续发展。为此，道路网络交通生命线优化设计的框架包括以下四个方面：

城市应急能力评估是道路网络交通生命线识别与优化设计的前提和基础。本研究提出考虑时变交通拥挤和人口分布的渐进合作应急可达性评估方法。渐进定义为应急需求点的可达性是与应急服务设施之间行驶时间的衰减函数；合作定义为多个应急设施可以协同对一个应急需求点的可达性产生影响。在评估过程中，该方法通过时变最短路径和动态人口数据考虑了交通拥挤和人口分布的时变特性。特别地，该方法不仅分析应急医疗、消防等单一类型应急服务的可达性，还探讨这些服务之间的相互作用与协同效应。

交通生命线识别是道路网络交通生命线优化设计的核心环节。本研究提出时变条件下基于路径冗余性识别交通生命线模型。该模型考虑OD（Origin-Destination）对、OD需求及交通拥挤的时变特征，通过计算各时段道路网络路径的冗余性，并结合时段权重来获得冗余性的期望值，从而科学地识别交通生命线。然而，由于需要考虑每个时段的路径冗余性，这给求解大规模现实城市道路网络的路径冗余性带来了计算挑战。为此，通过对城市网络结构进行重构，从而利用具有多项式计算时间性质的最大流和最小费用流算法，实现模型在有效时间内求解。

交通网络设计是道路网络交通生命线优化设计的具体实施阶段。本研究提出道路网络交通生命线路径冗余优化设计双层随机规划模型。上层目标是在有限的资金条件下，通过维护特定路段，来增强在各种可能灾害场景下的预期道路网络交通生命线路径冗余性；下层则针对每个特定的灾害场景，计算相应的预期道路网络交通生命线路径冗余性。该模型全面考虑需求和供给双方的不确定性。在需求方面，模型考虑不同灾害场景下OD对的变化带来的随机性；在供给方面，模型考虑不同灾害场景下道路交通拥挤的随机性。为了解决双层随机规划模型带来的计算挑战，本研究提出整数L型算法，该算法结合连续和整数最优割来分解模型，使得模型能够适用于现实的城市道路网络优化设计问题。

应急设施合理布置是道路网络交通生命线优化设计的重要补充。本研究提出道路网络交通生命线与应急设施布局综合优化设计两阶段随机规划模型。模型第一阶段关注交通生命线维护和应急设施选址的决策，而第二阶段则通过对每个潜在应急需求点分配最合适的应急设施，并确定在各种灾害场景下实现有效救援所需的最小响应时间。该模型充分考虑系统供需双方的随机特性。在需求方面，模型纳入灾害场景对应急服务需求量的影响。在供给方面，模型纳入灾害场景对道路网络拥堵的影响。此外，应急设施的容量也被视为随机变量，其大小取决于灾害事件的规模。鉴于考虑问题的复杂性，所得模型表现为非凸非线性混合整数规划，存在计算挑战。本研究利用线性化技术将原模型转化为一系列线性规划问题，并设计有效的不等式以加速算法的收敛，使得模型能够适用于现实的城市道路网络和应急设施布局问题。

为了检验上述道路网络交通生命线优化设计方法的有效性与系统性，本研究选取平陆运河钦州城区桥梁重建这一国家重大基础设施建设项目作为实证案例，取得了以下成果：一是验证了该方法在提升城市应急响应能力方面的有效性与高效性；二是展示了从应急能力评估到交通生命线识别、优化及应急设施布局的全链条系统串联性设计；三是解决了钦州城区在桥梁重建期间的交通平稳过渡与应急响应能力提升问题，为类似大型基础设施建设项目提供了宝贵的参考与借鉴。

关键词：道路网络,交通生命线,应急响应,随机规划,平陆运河

英文摘要

ABSTRACT

The urban road traffic network, as a core component of urban development, is an essential infrastructure for ensuring the smooth operation of daily urban activities. However, the network is often damaged by sudden events, severely affecting the efficiency of urban emergency responses and the safety of residents' lives and property. This study aims to establish a systematic methodology for optimizing urban road traffic lifelines to enhance rapid urban responses during emergencies and ensure urban safety and sustainable development. The framework for optimizing road traffic lifelines includes four key aspects:

Urban emergency capability assessment is the foundation for identifying and optimizing road traffic lifelines. This study proposes a gradual cooperative emergency accessibility evaluation method that considers time-varying traffic congestion and population distribution. The "gradual" aspect defines the accessibility of emergency demand points as a decay function of travel time to emergency service facilities, while the "cooperative" aspect defines the accessibility as being influenced by multiple emergency facilities working in coordination. This method incorporates time-varying shortest paths and dynamic population data to account for the changing characteristics of traffic congestion and population distribution. Specifically, it analyzes the accessibility of individual emergency services such as medical and fire services, while also exploring the interaction and synergy between these services.

Traffic lifeline identification is the core process in optimizing road traffic lifelines. This study proposes a time-varying traffic lifeline identification model based on path redundancy. The model considers the time-varying characteristics of origin-destination (OD) pairs, OD demand, and traffic congestion by calculating the redundancy of road network paths at different time intervals, combining these with the weights of each interval to obtain an expected value of redundancy, thus scientifically identifying the traffic lifelines. However, considering the redundancy of each time interval presents computational challenges when applied to large-scale urban road networks. To address this, the study reconstructs the urban network structure and utilizes maximum flow and minimum cost flow algorithms, which have polynomial time complexity, to solve the model within a feasible time frame.

Traffic network design represents the implementation phase of road traffic lifeline optimization. This study introduces a bi-level stochastic programming model for optimizing road traffic lifelines. The upper-level objective is to enhance the expected path redundancy of road traffic lifelines under various potential disaster scenarios by maintaining specific road links within a limited budget. The lower-level problem calculates the expected path redundancy for each disaster scenario. This model comprehensively considers uncertainties on both the demand and supply sides. On the demand side, it incorporates the randomness brought by changes in OD pairs under different disaster scenarios, while on the supply side, it considers the randomness of traffic congestion. To address the computational challenges posed by the bi-level stochastic programming model, this study proposes an integer L-shaped algorithm, which combines continuous and integer optimal cuts to decompose the model, making it suitable for real-world urban road network optimization problems.

Rational emergency facility placement is an essential supplement to optimizing road traffic lifelines. This study proposes a two-stage stochastic programming model that integrates road traffic lifelines with emergency facility layout optimization. The first stage focuses on decisions regarding the maintenance of traffic lifelines and the selection of emergency facility locations, while the second stage assigns the most appropriate emergency facility to each potential demand point and determines the minimum response time required for effective rescue under various disaster scenarios. The model fully considers the stochastic nature of both supply and demand. On the demand side, the model incorporates the impact of disaster scenarios on emergency service demand, while on the supply side, it considers the impact of disaster scenarios on road network congestion. Additionally, the capacity of emergency facilities is treated as a random variable, depending on the scale of the disaster. Given the complexity of the problem, the resulting model is a nonconvex nonlinear mixed-integer programming problem, posing computational challenges. This study employs linearization techniques to transform the original model into a series of linear programming problems and designs effective inequalities to accelerate algorithm convergence, making the model applicable to real-world urban road network and emergency facility layout problems.

To validate the effectiveness and systematic nature of the proposed road traffic lifeline optimization methodology, this study selects the bridge reconstruction project in Qinzhou city, part of the Pinglu Canal, a major national infrastructure project, as an empirical case. The following results were achieved: First, the method was proven effective and efficient in enhancing urban emergency response capabilities. Second, the study demonstrated a full-chain, systematic design linking emergency capability assessment, traffic lifeline identification, optimization, and emergency facility layout. Third, the study solved the problem of smooth traffic transitions and improved emergency response capabilities during the bridge reconstruction in Qinzhou city, providing valuable insights and references for similar large-scale infrastructure projects.

Key Words: Road Networks, Traffic Lifelines, Emergency Response, Stochastic Planning, Pinglu Canal

时玉琦

入学时间：2021级

答辩时间：2025年

论文题目：供给约束条件下城市道路信号控制交叉口交通设计方法研究

中文摘要

摘要

在城市建设由大规模增量扩张转向以存量提质增效为主的“城市更新”背景下，交通空间受限的老城区交叉口成为制约通行效率的关键瓶颈。此类交叉口具有道路宽度有限的几何特征，在高需求条件下容易出现空间供给与交通需求不匹配的问题。传统以新增车道为主的设计方式难以应对这种供给受限下的供需矛盾。针对这一问题，本文提出了面向供给约束条件下的城市道路信号控制交叉口交通设计方法，通过构建行为诊断模型、空间利用率提升模型和时空协调控制模型，实现对有限道路资源的精细化挖潜与动态适配，以提升交叉口的运行效率与资源利用水平。

交通设计作为交叉口时空资源分配的重要手段，长期以来被视为提升交通供给的核心方法。但传统的设计方法存在如下局限性：传统设计往往基于固定车道边界，忽视不同尺寸车辆对横向空间的差异化需求，导致车道横向空间利用效率不足；较少考虑多元交通主体的需求波动性，缺乏统一路权分配与动态复用设计，出现部分时段空间利用率高、而另一些时段资源空置浪费的问题；针对一些新型车道设计，未能有效解决需求的空间异质性与随机性相互作用下的时空资源浪费问题，使得新型交叉口的运行效率大打折扣；同时，由于难以量化设计要素和运行效果之间的关系，导致交通设计方案缺乏可解释性，且诊断流程多依赖人工观察和经验判断，效率低且主观性强。交通数据的丰富为进一步从更微观的供需适配层面进行交通设计、挖掘供给潜能提供了新的机遇。因此，本文从诊断入手，探索以存量挖掘为导向的精细化交叉口交通设计方法。

首先，针对交叉口缺乏诊断机制、宏观指标难以定位与解释根因，导致交通设计方案不能“对症下药”的问题，提出交叉口低效区域识别和致因诊断框架。具体而言，构建了基于深度最大熵逆强化学习的驾驶行为解析认知方法，并通过高斯混合模型识别交叉口低效区域，进而基于因果推断诊断低效的设计要素和控制策略等的致因，提出交叉口优化改善建议，形成诊断和设计的闭环。该方法以微观驾驶行为为基础、以因果识别为支撑，推动交通设计由传统经验驱动、仿真评价模式向数据驱动、可解释的智能化设计范式转变，为交叉口设计提供理论基础与方法支撑。

其次，针对需求的空间异质性特征导致的供给空间资源浪费问题，提出了新型多车道组合设计方法。具体而言，基于不同种类机动车横向尺寸的差异，提出了常规宽度车道、特殊宽度车道、小型车辆专用车道的组合设计方法，以最大化利用道路横向空间资源。首先，建立不同车道类型的通行能力模型，量化宽度缩减对运行效率的影响；随后，将车道数量、宽度与类型统一为决策变量，形成优化模型以确定典型交通条件下的最优车道布局，突破传统固定车道数量和统一车道宽度的假设。数值结果表明，相较传统车道设计，通过量平均提升了7.82%，且在车道数阶跃点附近优势显著。同时，适用性分析表明，在大型车辆占比低于80%时，该设计方法均高于传统车道设计方法，通过量普遍提升超过12%。

进一步，针对需求的多样性和时间波动性特征导致的供给空间资源在部分时段浪费问题，提出了统一路权分配与车道复用设计方法。具体而言，基于机动车和非机动车的需求特征，构建了双层随机规划模型，上层优化静态车道布局以确定车道边界，下层确定各时段车道的类型、功能与信号配时参数，以实现不同交通主体间的资源动态复用，有效应对非机动车流量在高峰时段激增的典型交通场景。数值结果表明，该设计方法可显著提升交叉口的通过量与资源利用效率，在高峰场景下相较静态方案提升9.90%、较非复用设计提升3.27%。同时，适用性分析表明，电动自行车占比由0%提升至100%时，通过量提升18.5%，但非机动车速度异质性对系统能力具有抑制作用。

最后，针对需求的空间异质性和时间随机性耦合特征导致的供给时空资源浪费问题，提出了面向新型车道设计的协同优化方法。具体而言，基于变结构车道设计下横向空间利用率上升的优势，进一步采用车道诱导方法提升纵向空间利用率，并结合自适应控制方案提升时间资源利用率。先构建了多车道潜在跟驰关系模型，再以平均延误最小化为目标识别换道车辆集合并确定车辆放行顺序，进一步以最小化车辆和停车线的距离为目标在安全约束下优化换道策略，最终通过滚动更新实现对需求波动的自适应。仿真结果表明，该方法在多种需求水平下均显著优于常规车道设计方案，通过量提升5.4%–15.5%，延误最高下降约50%；换道诱导策略在中高需求条件下效果最优；自适应配时有效缓解资源分配不均。同时，适用性分析表明，在大型车辆比例从0%增加至25%时，尽管整体通过量呈下降趋势，但所提方法始终优于传统设计与控制方案。

综上，本文针对多主体交通环境下的时空资源挖潜问题，先提出了基于诊断的设计闭环框架，在此基础上，进一步构建了面向供给空间资源利用率提升、供给空间资源在多元随机需求下的利用率提升、供给时空资源利用率提升的设计方法，引导交通设计走向精细化存量挖潜，具有重要的理论意义与实用价值。

关键词：城市交通，交叉口设计，供给约束，非常规车道，随机优化，时空协同建模

英文摘要

ABSTRACT

Against the backdrop of “urban renewal,” where city development is shifting from large-scale expansion to quality improvement of existing infrastructure, intersections in old urban areas with limited traffic space have become critical bottlenecks restricting traffic efficiency. These intersections are characterized by geometrically constrained road widths and are prone to mismatches between spatial supply and traffic demand under high-demand conditions. Traditional design approaches that rely primarily on adding new lanes are insufficient to address such supply–demand conflicts under constrained conditions. To tackle this problem, this study proposes a traffic design methodology for signalized intersections under supply constraints. By developing a behavior-based diagnostic model, a spatial utilization enhancement model, and a spatiotemporal coordination control model, the proposed framework aims to achieve refined utilization and dynamic adaptation of limited road resources, thereby improving intersection performance and resource efficiency.

Traffic design, as an important means of allocating spatiotemporal resources at intersections, has long been regarded as a core approach to enhancing transportation supply. However, conventional design methods have several limitations. They are often based on fixed lane boundaries, ignoring the differentiated lateral space requirements of vehicles of varying sizes, thereby resulting in low lateral space utilization. They seldom consider temporal fluctuations of multimodal traffic demand, lacking unified right-of-way allocation and dynamic lane reuse mechanisms, which causes high utilization in some periods and wasted capacity in others. For new lane configurations, existing studies fail to address the coupled effects of spatial heterogeneity and stochasticity of demand, leading to inefficient use of spatiotemporal resources and diminished operational performance at innovative intersections. Moreover, due to the difficulty of quantifying the causal relationship between design elements and operational outcomes, current designs lack interpretability, and diagnosis still relies heavily on manual observation and subjective experience. The growing availability of traffic data provides new opportunities to conduct traffic design from a more microscopic supply–demand matching perspective and to further explore latent capacity. Therefore, this study begins with diagnosis and develops refined intersection design methods oriented toward stock-based optimization.

First, to address the absence of diagnostic mechanisms and the inability of macroscopic indicators to localize and interpret inefficiencies—leading to “symptom-based” rather than “cause-based” design—this study proposes a framework for identifying low-efficiency regions and diagnosing causal factors at intersections. Specifically, a deep maximum-entropy inverse reinforcement learning approach is established to interpret driving behavior. Low-efficiency areas are then detected using Gaussian Mixture Models, and causal inference is employed to identify the underlying design and control factors responsible for inefficiency. Based on these findings, targeted design improvement strategies are proposed, forming a closed-loop between diagnosis and design. The framework leverages microscopic driving behavior and causal reasoning to advance traffic design from a traditional experience-driven and simulation-based paradigm toward a data-driven and interpretable intelligent design paradigm, providing a theoretical and methodological foundation for intersection optimization.

Second, to mitigate spatial supply inefficiency caused by the spatial heterogeneity of demand, a novel multi-lane combination design is proposed. Based on differences in the lateral dimensions of various vehicle types, the method introduces a combination of conventional width lanes, special width approach lanes, and dedicated passenger car lanes to maximize lateral space utilization. A capacity model for different lane types is first developed to quantify the impact of lane width reduction on performance. Then, the lane number, width, and type are unified as variables in an optimization model to determine the optimal lane layout under typical traffic conditions. Numerical results show that, compared with traditional lane designs, the proposed method increases throughput by an average of 7.82%, with particularly significant advantages near lane-number transition points. Applicability analysis further indicates that when the proportion of heavy vehicles is below 80%, the proposed design consistently outperforms conventional lane designs, with throughput improvements exceeding 12%.

Furthermore, to address temporal inefficiencies caused by demand diversity and time-varying patterns, a unified right-of-way allocation and lane-reuse design is developed. Based on the characteristics of motorized and non-motorized flows, a two-level stochastic programming model is constructed: the upper level optimizes static lane layouts to determine lane boundaries, while the lower level determines lane types, markings, and signal timing parameters for each time period. This enables dynamic resource reuse among traffic modes and effectively accommodates the surge of non-motorized traffic during peak periods. Numerical experiments demonstrate that the proposed method significantly enhances throughput and resource utilization. Under peak conditions, throughput increases by 9.90% compared with static designs and by 3.27% compared with non-reuse schemes. Applicability analysis reveals that as the proportion of electric bicycles rises from 0% to 100%, overall throughput improves by 18.5%, though excessive heterogeneity in non-motorized speeds suppresses system capacity.

Finally, to address spatiotemporal inefficiencies resulting from the coupled spatial heterogeneity and temporal randomness of demand, a collaborative optimization method for innovative lane designs is proposed. Building on the improved lateral utilization of special width approach lane, the method integrates lane-guidance strategies to enhance longitudinal utilization and adaptive signal control to optimize temporal efficiency. A potential multi-lane car-following model is developed, followed by identifying candidate lane-changing vehicles and optimizing their release sequences to minimize average delay. Subsequently, a safety-constrained optimization model is formulated to minimize the distance between vehicles and the stop line during lane changes, and a rolling-update mechanism is introduced to adapt to demand fluctuations. Simulation results show that the proposed approach outperforms conventional designs under various demand levels, improving throughput by 5.4%–15.5% and reducing delay by up to 50%. Lane-guidance strategies are most effective under medium to high demand, and adaptive signal control further mitigates uneven resource allocation. Applicability analysis indicates that as the proportion of heavy vehicles increases from 0% to 25%, overall throughput declines, yet the proposed method consistently surpasses traditional design and control approaches.

In summary, this study addresses the problem of spatiotemporal resource optimization under various traffic conditions by first proposing a diagnosis–design closed-loop framework. Building upon this foundation, three refined design methodologies are developed to enhance spatial resource utilization, utilization under multi-modal stochastic demand, and overall spatiotemporal efficiency. Together, these contributions advance intersection design toward refined stock-based optimization, offering substantial theoretical significance and practical value for supply-constrained urban transportation systems.

Key Words: Urban Traffic; Intersection Design, Constrained Supply, Novel Lanes, Stochastic Optimization, Spatiotemporal Collaborative Modeling