A game theory-based approach for modelling mandatory lane-changing behaviour in a connected environment

A game theory-based approach for modelling mandatory lane-changing behaviour in a connected environment
复制标题

DOI:
10.1016/j.trc.2019.07.011
复制
发表时间:
2019-09-01
影响因子:
8.3
通讯作者:
Wang, Meng
Wang, Meng
中科院分区:
工程技术1区
文献类型:
--
作者:
Ali, Yasir;Zheng, Zuduo;Wang, Meng

文献摘要

被引文献

相似文献

联网环境提供有关周围交通的实时信息;这些信息有助于复杂的驾驶操作,例如需要有关周围车辆信息的变道。到目前为止,连接环境中的变道建模很少受到关注。这是由于连接环境的新奇以及随之而来的数据稀缺性。一个行为上合理的车道变换模型甚至不适用于传统的环境,也就是说,一个没有驾驶辅助的环境。为了满足这一需求,本研究开发了一个基于博弈论的强制换道模型(AZHW模型)的传统环境,并将其扩展到连接的环境。CARRS-Q高级驾驶模拟器用于为联网环境收集高质量的车辆轨迹数据。开发的模型(传统环境和连接环境)使用NGSIM和模拟器数据在一个双层校准框架中进行校准。使用各种性能指标对模型的性能进行了严格的评估。这些包括真阳性、假阳性、检测率、虚警率、时间预测误差和位置预测误差。结果一致表明,开发的基于博弈论的模型可以有效地捕捉强制换道的决定,具有很高的准确度。此外,开发的AZHW模型的性能进行了比较,在文献中的代表性的基于博弈论的换道模型。对比分析表明,本研究中开发的AZHW模型优于现有的模型。
The connected environment provides real-time information about surrounding traffic; such information can be helpful in complex driving manoeuvres, such as lane-changing, that require information about surrounding vehicles. Lane-changing modelling in the connected environment has so far received little attention. This is due to the novelty of connected environment, and the consequent scarcity of data. A behaviourally sound lane-changing model is not even available for the traditional environment; that is, an environment without driving aids. To address this need, this study develops a game theory-based mandatory lane-changing model (AZHW model) for the traditional environment and extends it for the connected environment. The CARRS-Q advanced driving simulator is used to collect high-quality vehicle trajectory data for the connected environment. The developed models (for traditional environment and connected environment) are calibrated using NGSIM and simulator data in a bi-level calibration framework. The performance of the models has been rigorously evaluated using various performance indicators. These include the true positive, false positive, detection rate, false alarm rate, time prediction error, and location prediction error. Results consistently show that the developed game theory-based models can effectively capture mandatory lane-changing decisions with a high degree of accuracy. Furthermore, the performance of the developed AZHW models is compared with representative game theory-based lane-changing models in the literature. The comparative analysis reveals that the AZHW models developed in this study outperform existing models.