考虑驾驶风格的主从博弈决策规划算法研究

Research on Decision-Making and Planning Algorithms for Stackelberg Games Considering Driving Styles

摘要: 在混合驾驶的高速公路场景中，自动驾驶车辆与网联人类驾驶车辆均有着不同的驾驶风格，如何合理建模其交互以提升决策规划的安全性，是自动驾驶的关键问题之一。针对上述问题，本文提出了一种考虑车辆驾驶风格的主从博弈决策规划算法。首先，使用K-means聚类法将车辆驾驶风格划分为激进型、正常型、保守型三类。其次，引入主从博弈对自车与网联人驾车间的交互进行建模，并基于高斯分布对他车运动不确定性进行量化，从而求解出决策最优解。最后，轨迹规划模块利用该解，通过路径规划和速度规划解耦的方法生成轨迹规划结果。PreScan、Matlab和CarSim联合的仿真实验结果表明，所提出的决策规划算法能有效应对具有不同驾驶风格的车辆间的交互，从而得出安全可靠的轨迹。

Abstract: In mixed traffic highway scenarios, autonomous vehicles and connected human-driven vehicles exhibit different driving styles, and how to reasonably model their interactions to improve the safety of decision-making and planning remains one of the key challenges in autonomous driving. To address these issues, this paper proposes a Stackelberg game-based decision-making and planning method that takes vehicle driving styles into account. Firstly, the K-means clustering method is used to classify vehicle driving styles into three categories: aggressive, normal, and conservative. Then, the Stackelberg game is introduced to model the interaction conflicts between autonomous vehicles and human-driven vehicles, and the motion uncertainty of other vehicles is quantified based on Gaussian distribution to solve for the optimal decision. Finally, the planned trajectory is generated in the trajectory planning module by utilizing the optimal decision and applying a decoupled approach to path and speed planning. The simulation experiment results based on the integrated PreScan, Matlab and CarSim platform demonstrate that the proposed decision-making and planning method can effectively handle interactions between autonomous vehicles and human-driven vehicles with different driving styles.