基于分子力场的网联自主车辆跟驰安全特性及模型

Car-following Model and Safety Characteristics of Connected Autonomous Vehicle Based on Molecular Force Field

为更精准地描述人车路复杂环境下网联异质车流多维感知与交互所呈现的动态特性,本文提出运用分子动力学势场函数定量描述车辆间相互作用关系的车辆跟驰行为模型,从分子力场角度剖析网联自主车辆的自驱动粒子特性和车车交互行为安全特性,独辟系统分析网联异质车群协同关系和安全态势演化规律的新思路。首先,类比复杂交通场景的网联自主车辆为自驱动粒子,基于分子动力学势场理论推导出网联自主车辆跟驰行为模型表达式,引入速度协同项,建立网联自主车辆跟驰行为的分子力场模型;其次,运用人工蜂群算法,采用High_D车辆轨迹数据对已有的跟驰模型和智能驾驶员模型进行参数标定,验证分析分子力场跟驰模型的合理性及安全性;设计数值仿真实验检验分子力场模型对真实车辆跟驰行为的拟合效果及稳定性表现。数值仿真结果表明,分子力场模型得出的车辆加速度结果与实际数据的平均值绝对误差与均方根误差更低,且受到扰动时波动更小,该模型提升了网联自主车辆跟随行驶的安全和效率,宏观车流运行具有更好的稳定性。网联自主车辆的分子力场跟驰模型可系统描述网联异质车群的动态特性、微观车辆跟驰行为和车车安全交互作用关系,为提高行车安全性及通行效率提供理论基础。

To describe the dynamic characteristics of multi-dimensional perception and interaction of heterogeneous traffic flow under the complex environment of human-vehicle-road more accurately,this paper proposes a vehiclefollowing model based on the molecular dynamics potential field function.The characteristics of self-driven particles and the safety characteristics of vehicle to vehicle interaction behavior of connected autonomous vehicles are analyzed from the perspective of molecular force field,which is helpful for systematically analyzing the synergy relationship and safety situation evolution law of networked heterogeneous vehicle groups.First,the connected autonomous vehicles under complex traffic conditions are taken as self-driven particles.The vehicle-following model for the connected autonomous vehicle is established based on the molecular dynamics potential field theory.The molecular force field model for the following behavior of the connected autonomous vehicles is developed by introducing the velocity synergy term.Then,the Artificial Bee Colony Algorithm is used to calibrate the parameters of the existing carfollowing model and the intelligent driver model using the High_D vehicle trajectory data.The rationality and safety of the molecular force field car-following model are verified.The numerical simulation is designed to verify the fitting effect and stability performance of the molecular force field model on the real vehicle following behavior.The results show that the Mean Absolute Error and Root Mean Squared Error of the vehicle acceleration results obtained by the molecular force field model were lower and the fluctuation was smaller than actual data when disturbed.The proposed model improves the safety and efficiency of the following behavior of the connected autonomous vehicles,and the macroscopic traffic flow operation has better stability.The proposed model can systematically describe the dynamic characteristics,microscopic car-following behavior and the vehicle to vehicle safety interaction relationship of connected heterogeneous vehicle groups,which provides a theoretical basis for improving driving safety and traffic operational efficiency.