Abstract

Connected and automated vehicles (CAVs) are expected to change the way we travel. Before both the vehicles and infrastructures are fully automated, users of CAVs are required to respond appropriately to any adverse on-road conditions or malfunction that may prevent the autonomous driving system from reliably sustaining the dynamic driving task performance. The objective of this study is to construct spatiotemporal driving volatility profiles to help CAVs or drivers identify the potential hazards in the existing transportation network and make proactive driving decisions. The volatility profiles are constructed based on the historical traffic dynamics, varying spatially and temporally in the network. For demonstration, this study exploited the Basic Safety Messages datasets from Safety Pilot Model Development program in Ann Arbor, Michigan. The driving volatility is a measure to reflect the variability of driving performance, which is often used to show a vehicle or driver’s performance on road. This study extends the concept to capture the driving dynamics as a performance of the transportation network. This study also matched the driving volatility to the spatial and temporal occurrence of historical traffic crashes. Modeling results showed the volatility is significantly related to safety outcomes; therefore, the driving volatility profiles can be compiled into the high definition (HD) maps to inform CAVs and drivers of potential on-road hazards and assisting in making proactive driving decisions. Further, the results offer implications for potential upgrades of the transportation infrastructure for full automation in the future.

摘要

联网和自动驾驶汽车 (CAV) 有望改变我们的出行方式。在车辆和基础设施完全自动化之前，CAV 的用户需要对可能阻止自动驾驶系统可靠地维持动态驾驶任务性能的任何不利的道路条件或故障做出适当的响应。本研究的目的是构建时空驾驶波动曲线，以帮助 CAV 或驾驶员识别现有交通网络中的潜在危险并做出主动驾驶决策。波动性剖面是基于历史流量动态构建的，在网络中随空间和时间变化。为了演示，本研究利用了密歇根州安娜堡安全试点模型开发计划中的基本安全消息数据集。驾驶波动性是反映驾驶性能变化的一种量度，常用于表示车辆或驾驶员在道路上的表现。这项研究扩展了这一概念，将驾驶动态作为交通网络的一种表现。该研究还将驾驶波动性与历史交通事故的时空发生率相匹配。建模结果表明，波动性与安全结果显着相关；因此，驾驶波动曲线可以编译到高清 (HD) 地图中，以告知 CAV 和驾驶员潜在的道路危险，并协助做出积极的驾驶决策。此外，研究结果为未来交通基础设施的全面自动化升级提供了启示。