摘要
为突破车辆自适应巡航控制系统对行驶速度、道路线形和交通流状态的运行设计域限制,提出一种由类人决策层与协同控制层组成的双层弯道自适应巡航控制(Bilevel Curve Adaptive Cruising Control,B-CACC)策略。首先,根据驾驶人弯道视觉特性构建风险注意分布(Risk Attention Distribution,RAD)模型,模拟预瞄距离伸缩与注意力动态特性。基于中枢能量理论构建驾驶负荷分配(Driving Load Distribution,DLD)模型,以刺激与意愿传递关系解耦人-车-路协同下的横纵向运动。以此为基础,结合智能驾驶人模型提出弯道自适应巡航决策方法。其次,运用三自由度动力学和运动学模型,通过泰勒级数展开与前向欧拉法开发模型预测控制算法。以类人决策约束替代状态量与控制量约束,运用数值求解方法滚动优化转向控制。采用比例积分控制车辆纵向运动,构建车辆横纵向协同的轨迹跟踪控制器。最后,应用Carsim与MATLAB/Simulink搭建模型在环的联合仿真试验,分别在自由与跟驰状态下测试RAD与DLD模型作用效果,检验所提B-CACC策略的有效性。试验结果表明:RAD模型作用下车速和曲率的相对分布符合驾驶人特性,保障了侧向加速度期望与输出的一致性;DLD模型以空间换时间,有效缓减了跟车时的运动冲击,在类人决策作用下,横向轨迹跟踪的最大绝对误差和均方根误差分别降低31.4%和21.4%,提高了B-CACC对曲率变化与前车干扰的稳定性与鲁棒性,改善了巡航安全性与舒适性。
To overcome the operational design domain limitations of vehicle-adaptive cruise control systems in terms of the travel speed,road alignment,and traffic flow state,a bilevel curve adaptive cruise control(B-CACC)strategy composed of a human-like decision-making layer and cooperative control layer was developed.First,a risk attention distribution(RAD)model was constructed based on a driver's visual characteristics at curves to simulate preview distance stretching and the dynamic characteristics of attention.Based on the central capacity theory,a driving load distribution(DLD)model was constructed to decouple the lateral and longitudinal motions of human-vehicle-road coordination under the relationship between the stimulus and willingness transmission.Accordingly,an adaptive cruise decision-making method for curves was proposed and combined with an intelligent driver model.Second,a model predictive control algorithm was developed using Taylor series expansion and the forward Euler method using three-degree-of-freedom dynamic and kinematic models.By replacing the state and control quantity constraints with human-like decision constraints,a numerical solution method was applied to the rolling optimization of steering control.The trajectory tracking controller for the vehicle's lateral and longitudinal coordination was constructed based on the proportional integral control of the vehicle's longitudinal motion.Finally,CarSim and MATLAB/Simulink were used to conduct a model-in-the-loop co-simulation test.The effects of the proposed RAD and DLD models in the free and following states were determined to verify the effectiveness of the proposed B-CACC strategy.The experimental results showed that the relative distributions of the vehicle speed and curvature under the RAD model were consistent with the driver characteristics,which ensured the consistency of the lateral acceleration expectation and real output.The DLD model effectively reduced the impact of motion when following a vehicle by making changes in space over time.Under the action of the human-like decision-making,the maximum absolute error and root mean square error of the lateral trajectory tracking were reduced by 31.4%and 21.4%,respectively.Human-like decision-making improved the stability and robustness of the B-CACC against the interference of curvature change and the vehicle in front,and accordingly improved the cruise safety and comfort.
作者
韩天园
沈永俊
鲍琼
屈琦凯
吴臻
HAN Tian-yuan;SHEN Yong-jun;BAO Qiong;QU Qi-kai;WU Zhen(School of Transportation,Southeast University,Nanjing 211189,Jiangsu,China)
出处
《中国公路学报》
EI
CAS
CSCD
北大核心
2023年第10期211-223,共13页
China Journal of Highway and Transport
基金
国家自然科学基金项目(52002063)
教育部人文社科基金项目(21YJCZH129)。
关键词
汽车工程
弯道巡航
类人决策
风险注意分布
驾驶负荷分配
模型预测控制
横纵向协同
自动驾驶
automotive engineering
curve cruising
human-like decision-making
risk attention distribution
driving load distribution
model predictive control
lateral-longitudinal coordination
automated driving