基于路段的风险型交通分配模型与网络风险评估方法

Travel time uncertainty has a significant impact on transportation systems. Travelers and planners treat travel time uncertainty as a risk in their travel choices and planning decisions. A systematic risk assessment of transportation networks is critical in the capability enhancement of hedging against uncertainty. The existing methodology of evaluating network performance under uncertainty is mainly based on the concept of reliability, which accounts for the occurrence probability of risk. However, it is incapable of capturing the potential loss due to risk. This project will address the following two fundamental questions: how to reasonably model travelers’ individual risk-averse route choice behaviors and the aggregated link flow pattern while considering the congestion effect, and how to systematically characterize the network risk under uncertainty. First of all, we will use the concept of mean-excess travel time to integrate the magnitude and the occurrence probability of the potential risk in travelers’ route choice decisions. Based on this route choice criterion, we will develop a link-based risk-averse traffic assignment model with a particular consideration of large-scale network applications. The network-wide performance uncertainty is not just a simple sum of the link-level uncertainty. Instead, it involves the spatial propagation and aggregation of link-level uncertainty. To deal with this complexity, a high-order moment method will be developed to quantify the multi-dimensional characteristics in the uncertainty propagation, such as the asymmetry and peakedness. Also, we will adopt the concept of copula to capture different types of link dependency structure in the uncertainty aggregation. With the behavior modeling and uncertainty propagation, a novel network-wide risk measure will be developed to smartly integrate the magnitude and the occurrence probability of the potential risk in network performance. This risk measure is able to capture both the acceptable risk within the planners’ reliability requirement and the unacceptable risk of exceeding the reliability requirement. To cater for the multi-objective requirements, such as efficiency, energy consumption, and emission, the proposed risk measure will be further applied to different performance dimensions. To enhance the practicality of the risk measure, we will develop an analytical estimation method to directly calculate the risk measure without fitting the explicit distribution form of network performance uncertainty. Finally, a real case study will be conducted to validate the proposed traffic assignment model and network-wide risk assessment method. The products of this research will provide a methodological foundation for systematically evaluating the network performance under uncertainty, and also enhance the realism of network evaluation and resource allocation.

行程时间不确定性对交通系统有着显著影响，出行者和规划者将之视为风险。全面评估交通网络中的风险是提升网络应对不确定性能力的关键。现有基于可靠度的网络性能评价仅考虑风险出现的概率，而未兼顾风险可能带来的损失。本项目研究如何合理描述出行者的风险规避路径选择行为和网络流量均衡状态，如何全面剖析不确定环境下网络风险的特征。在综合考虑风险概率及损失程度的基础上，研究基于均值-超量的路径选择行为，建立适用于大规模路网的路段风险型交通分配模型；以高阶矩和copula为方法，刻画不确定性由供需源头到路段流量再集计到网络性能的传播过程；建立多目标性能的均值-超量网络风险指标及解析型估计方法，全面捕捉网络性能的可接受风险和不可接受风险，有机整合风险概率和损失程度。基于大规模路网，开展实证分析，验证模型和方法的有效性。研究成果可为系统全面地评估不确定交通网络的性能提供方法论基础，提升网络评价和资源配置的科学性。

交通系统受到许多不确定性因素的影响，如交通需求的波动、路段容量的退化，这些不确定性都会导致行程时间的不确定性。实证研究表明，行程时间不确定性对出行者的行为决策和交通系统的性能有着显著影响。全面评估交通网络中的风险是提升网络应对不确定性能力的关键。本项目试图构建不确定环境下交通网络性能评估的系统化理论体系和技术方法，包括不确定环境下的交通分配模型与算法、交通网络冗余性和脆弱性评估方法、不确定性对网络检测器布局优化和建设项目比选的影响研究。具体而言，针对客观的供需不确定性，建立了不确定环境下基于路段的均值-超量交通分配模型，克服了现有路径型模型在建模灵活性、面向大规模网络应用的算法设计方面的缺憾；针对主观不确定性(即出行者的随机感知误差)，建立了能够同时刻画绝对和相对费用差的解析型路径选择行为模型及其等价的随机用户均衡交通分配模型，同时克服了现有广泛使用的Logit模型和Weibit模型的缺陷；提出了交通网络冗余性的系统化分析框架(包括定义、评价指标、计算方法、优化策略等)，为交通网络评价与优化提供了全新的分析维度，为提升网络应对灾害的恢复力提供了方法论基础；在此基础上，提出了交通网络脆弱性包络线的概念，建立了多路段同时失效情形下交通网络脆弱性的上下界模型，可有效支撑交通网络关键路段识别和资源配置；利用网络流量完整可观测性条件和线性规划理论，建立了检测器布局鲁棒优化模型，对输入数据要求低，可应用于大规模交通网络；以BOT交通项目比选为例，在理论上从不同的角度证明了不确定性所造成的基于最大收益或最低成本准则的比选结果具有内在的选择性偏见，它可能是交通需求预测结果不准确的另一内因，并设计了消除偏见的策略。研究成果丰富了交通网络流理论体系，可为系统全面地评估不确定交通网络的性能提供方法论基础，增强了满足精细化交通系统分析需求的能力，提升了交通网络规划的鲁棒性和可靠性。