Assessing Crash Risks of Evacuation Traffic: A Simulation-based Approach

Recently, hurricanes have caused major concern for transportation agencies and policymakers attempting to find better evacuation strategies. This was especially evident after Hurricane Irma, which forced about 6.5 million Floridians to evacuate the state. This mass evacuation caused a significant amount of delays on state highways due to heavy congestion and car crashes. Crashes and accidents on roads and highways are of major concern during evacuation efforts. Though several strategies have been implemented to manage the heavy traffic demand during a hurricane evacuation, current approaches seem to have less of an impact on traffic safety. In this context, this project had three objectives:
To assess the impact of hurricane evacuation on crash risks,
To identify if there are any changes in traffic flow behavior between evacuation and non-evacuation periods, and
To assess the impact of an in-vehicle driving assistance system during an evacuation period.
First, to assess the impact of hurricane evacuation on crash risks, we adopted a matched case control approach. After collecting traffic and crash data along a major evacuation route in Florida, we estimated models for three different conditions: regular period, evacuation period, and a combination of both evacuation and regular period data. Model results show that if there is high occupancy at an upstream station and high variation of speed at a downstream station, the probability of crash occurrence increases. We estimate the effect of evacuation itself on crash risk and find that, after controlling for traffic characteristics, during evacuation the chance of an accident is higher than in a regular period. These findings will help us develop advanced real-time crash prediction models which will work for evacuation traffic conditions, and design proactive countermeasures to reduce crash occurrences during evacuation.
Second, to understand driver behavior during evacuation and to assess the potential safety impacts of adaptive cruise control (ACC) systems, we developed a microscopic simulation model in SUMO for a segment of the Interstate highway 75 (I-75), and calibrate it using real-world traffic data collected from the evacuation period of hurricane Irma. For the calibrated model, we find that the values of maximum acceleration and deceleration are 4.5 m/s^2 and 6.5 m/s^2, respectively. These values are higher than those in typical car-following models calibrated under regular traffic conditions. Also, higher acceleration and deceleration values indicate abrupt speed variation, which is the most common scenario for evacuation traffic. To evaluate the safety impact of ACC systems, we adopted two surrogate measures: time to collision (TTC) and deceleration rate to avoid a collision (DRAC). Our experiment results show that during evacuation, about 49% of traffic collisions can be reduced at a 25% market penetration of ACC-equipped vehicles.
The findings from this project have further implications for evacuation declarations and highlight the need for better traffic management strategies during evacuation. Based on the findings, we propose several traffic management strategies to reduce the number of crashes during evacuation. We also propose solutions based on in-vehicle driving assistance systems and identify the challenges to increase market penetration rate for such technologies.

近年来，飓风已成为交通管理机构与政策制定者探索更优疏散策略时的重点关切问题。这一点在飓风艾尔玛（Hurricane Irma）事件中体现得尤为突出：该飓风迫使佛罗里达州约650万民众撤离本州。此次大规模疏散行动因严重拥堵与车辆碰撞事故，导致州内公路出现大面积延误。疏散期间道路与高速上的碰撞事故始终是核心关切点。尽管已有多项策略用于应对飓风疏散期间的高强度交通需求，但现有方案对交通安全的改善效果仍不尽如人意。在此背景下，本项目确立了三大目标：
1. 评估飓风疏散行动对碰撞风险的影响；
2. 识别疏散与非疏散时段交通流行为是否存在差异；
3. 评估车载驾驶辅助系统在疏散时段的应用效果。

首先，为评估飓风疏散对碰撞风险的影响，本研究采用匹配病例对照研究方法。我们收集了佛罗里达州一条主要疏散通道的交通与碰撞数据，随后针对三种场景分别构建模型：常规通行时段、疏散时段，以及疏散与常规时段的混合数据场景。模型结果显示，若上游监测站点的车辆占有率较高，且下游监测站点的车速波动幅度较大，则碰撞发生概率会显著提升。同时，在控制交通特征变量后，本研究估算得出疏散时段的事故发生概率高于常规通行时段。上述研究结果将助力开发适用于疏散交通场景的先进实时碰撞预测模型，并设计主动干预措施以降低疏散期间的碰撞事故发生率。

其次，为理解疏散期间的驾驶员行为并评估自适应巡航控制系统（Adaptive Cruise Control, ACC）的潜在安全效益，本研究基于SUMO（Simulation of Urban MObility）仿真平台，针对75号州际公路（Interstate Highway 75, I-75）的一段路段构建了微观交通仿真模型，并利用飓风艾尔玛疏散时段采集的真实交通数据完成模型校准。校准后的模型参数显示，车辆最大加速度与最大减速度分别为4.5 m/s²与6.5 m/s²，该数值高于常规交通场景下校准得到的跟驰模型参数。此外，更高的加减速幅值意味着更剧烈的车速变化，这也是疏散时段交通的典型特征。为评估ACC系统的安全效益，本研究采用两项替代评价指标：碰撞时间（Time to Collision, TTC）与避免碰撞所需减速度（Deceleration Rate to Avoid a Collision, DRAC）。仿真实验结果表明，在疏散时段，当配备ACC的车辆市场渗透率达到25%时，约49%的交通碰撞事件可被有效降低。

本项目的研究结果对飓风疏散预案制定具有重要参考价值，同时凸显了优化疏散期间交通管理策略的必要性。基于上述发现，本研究提出了多项可降低疏散期间碰撞事故数量的交通管理方案，并围绕车载驾驶辅助系统提出了落地解决方案，同时指出了提升该类技术市场渗透率所面临的挑战。