This paper presents a hierarchical and game-theoretic decision-making strategy for connected and automated vehicles (CAVs). A CAV can receive preview information using vehicle-to-everything (V2X) communication systems, and the optimal short- and long-term trajectory can be planned using this information. Specifically, in this study, the aggressiveness of all preceding vehicles in the car-following scenario can be estimated globally by monitoring the history of their time-series behaviors, before the CAV initiates a particular action, which is performed at the upper layer of the proposed decision-making structure. If it is determined that initiating a specific action is advantageous, the action is initiated, and the CAV then interacts with the vehicles locally to achieve its driving goal in a game-theoretical manner at the lower layer. In multiple test scenarios, we demonstrate the usefulness of our approach compared to the conventional decision-making approaches, and it shows a significant improvement in terms of success rates.

本文提出了一种用于联网和自动驾驶汽车 (CAV) 的分层和博弈论决策策略。CAV 可以使用车联网 (V2X) 通信系统接收预览信息，并可以使用此信息规划最佳的短期和长期轨迹。具体来说，在这项研究中，在 CAV 启动特定动作之前，可以通过监视其时间序列行为的历史来全局估计跟车场景中所有前方车辆的攻击性，该动作在网络的上层执行拟议的决策结构。如果确定启动特定动作是有利的，则启动该动作，然后CAV在本地与车辆交互以在下层以博弈论方式实现其驾驶目标。