基于改进行车风险场的自动驾驶汽车轨迹规划研究

Research on Trajectory Planning for Autonomous Vehicles Based on Improved Driving Risk Field

摘要: 在复杂的交通环境中，准确建立行车风险模型并量化风险分布，是评估自动驾驶安全性的关键，也是系统实现类人决策的基础。然而，现有的风险场模型在处理多种风险要素耦合以及适应动态场景时仍存在不足。为此，本研究提出了一种融合多源风险要素的改进型行车风险场模型，分别构建了动态交互风险场与静态约束风险场，成功将行车风险转化为可以量化的空间分布场强。随后，本研究将该模型深度嵌入到自动驾驶的换道决策框架中。系统能够在决策层完成实时的风险评估与路径优化，从而有效应对复杂场景下的换道需求。最后，仿真实验结果表明，该改进模型能够显著降低车辆的碰撞风险概率，并有效提升行驶效率。本研究为自动驾驶的决策与规划提供了一种兼具安全性与鲁棒性的解决方案。

Abstract: In complex traffic environments, accurately establishing a driving risk model and quantifying risk distribution is crucial for assessing the safety of autonomous driving, and it also serves as the foundation for the system to achieve human-like decision-making. However, existing risk field models still have deficiencies in handling the coupling of multiple risk factors and adapting to dynamic scenarios. To address this, this study proposes an improved driving risk field model that integrates multi-source risk factors. This model constructs a dynamic interactive risk field and a static constrained risk field, successfully transforming driving risks into quantifiable spatial distribution field intensities. Subsequently, this study deeply embeds this model into the lane-changing decision-making framework of autonomous driving. The system is capable of completing real-time risk assessment and path optimization at the decision-making level, effectively addressing lane-changing demands in complex scenarios. Finally, simulation experimental results show that the improved model can significantly reduce the probability of vehicle collisions and effectively enhance driving efficiency. This study provides a solution for autonomous driving decision-making and planning that combines safety and robustness.