Collaborative Research: Understanding the Impacts of Automated Vehicles on Traffic Flow Using Empirical Data

新兴的自动驾驶汽车（AV）技术可能会颠覆和改变我们的交通系统。关于自动驾驶汽车的大量研究都是基于自动驾驶汽车相对于其他车辆的行为假设。不幸的是，由于av的缺失，很少有假设可以得到经验验证。然而，自动驾驶技术的一个关键组成部分，自适应巡航控制（ACC），已经使用了十多年，可以用来填补这一空白。该研究旨在研究具有ACC的车辆在与道路上的其他车辆互动时的行为。这项研究将为未来自动驾驶汽车的行为提供重要的见解。从这项研究中获得的理解也将对交通管理、交通规划、ACC车辆和自动驾驶汽车的设计产生重要影响。此外，该项目将参与一系列综合研究、教育和推广活动，包括与研究和实践社区分享ACC数据、开发教育模块，以及通过夏令营进行K-12推广。更具体地说，研究将(i)收集不同汽车制造商的ACC车辆及其对应的普通车辆（rv）的经验轨迹数据；（ii）制定捕捉ACC和rv相似和不同特征的汽车跟随模型。该项目将重点关注使用最大似然估计（MLE）的数据收集和模型估计工作。这使得统计推断方法（例如，似然比检验）的新应用能够检验各种假设，以评估不同ACC系统之间以及ACC和rv之间的差异和相似性。特别是，这项研究将测试不同汽车制造商的ACC系统是否彼此不同，是否与房车不同，以及它们是否随着时间的推移而发生重大变化。了解这一点很重要，因为它将决定未来该领域的研究是否必须专注于分析每个汽车制造商，或者ACC系统最终是否会向类似人类驾驶的方向发展。在建模工作中，研究将建立一个通用的随机模型，使ACC车辆和rv的行为能够协调一致。该研究将检查模型组件分布的不同变化来指定模型，并使用最大似然估计模型参数。此外，研究将使用测试的ACCs的估计模型将结果外推到一般混合交通。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Emerging automated vehicle (AV) technologies are likely to disrupt and transform our transportation system. The vast number of studies on AV hinge upon assumptions on how AVs behave with respect to other vehicles. Unfortunately, few of the assumptions can be empirically validated due to the absence of AVs. And yet, a critical component of AV technologies, the adaptive cruise control (ACC), has been used for over a decade and can be used to fill this gap. The research aims to study how vehicles with ACC behave when interacting with other vehicles on the road. The research will provide important insights on the behaviors of AVs in the future. The understandings gained from this research will also have important implications in traffic management, transportation planning, and design of ACC vehicles and AV. Additionally, this project will engage in a range of integrated research, educational and outreach activities, including sharing the ACC data with the research and practice community, developing educational modules, and K-12 outreach through summer camp. More specifically, the research will (i) collect empirical trajectory data of ACC vehicles of different car-makers and their counterparts, regular vehicles (RVs), and (ii) formulate car-following models that capture the similar and differentiating features of ACCs and RVs. The project will focus on data collection and model estimation efforts using Maximum Likelihood Estimation (MLE). This enables the novel applications of statistical inference methods (e.g., the likelihood ratio test) to test various hypotheses to assess the differences and similarities among different ACC systems and between ACCs and RVs. In particular, the research will test if ACC systems from different car-makers differ from one another and from the RVs, and if they change substantially over time. Knowing this is important because it will dictate whether or not future research in this area has to focus on analyzing each individual carmaker, or if ACC systems will eventually converge towards human-like driving. In the modeling efforts, the research will build a general stochastic model so that the behaviors of ACC vehicles and RVs can be reconciled. The research will examine different variations of the distributions of the model components to specify the model(s) and use MLE for estimation of model parameters. Additionally, the research will use the estimated model(s) of the tested ACCs to extrapolate the results to the general mixed traffic.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.