基于运动学原理的换道轨迹预测  

Lane Changing Trajectory Prediction Based on Kinematics Principle

作　　者：刘磊[1] 王易[1] 康凯[1] 李洪庆 

机构地区：[1]中国汽车技术研究中心有限公司,天津

出　　处：《交通技术》2022年第3期260-274,共15页Open Journal of Transportation Technologies

摘　　要：换道作为车辆的基本驾驶行为之一,相较于跟驰行为,换道过程更具复杂性,对道路上车辆的运行安全有着至关重要的影响。为了对车辆换道轨迹进行准确预测,满足自动驾驶条件下车辆换道的安全性与合理性,本文选定五次多项式为换道轨迹预测基础模型,构建效益函数和选取约束条件,应用MATLAB中的fmincon优化工具箱,使用二次序列优化算法对模型参数进行寻优求解。对比换道预测轨迹与实际轨迹,对模型的有效性进行验证,结果表明预测轨迹的横向位移绝对误差集中在−0.3 m~0.3 m之间,偏转角速度波动峰值均小于2˚/s,加速度维持在0~2 m/s2的范围内,满足轨迹合理性、换道舒适性和平稳性。Lane changing is one of the basic driving behaviors of vehicles. Compared with car following behavior, the lane changing process is more complex and has a vital impact on the operation safety of vehicles on the road. In order to accurately predict the vehicle lane change trajectory and meet the safety and rationality of vehicle lane change under the condition of automatic driving, this paper selects the quintic polynomial as the basic model of lane change trajectory prediction, constructs the benefit function and selects the constraints, applies the fmincon optimization tool box in MATLAB, and uses the quadratic sequence optimization algorithm to optimize the model parameters. The effectiveness of the model is verified by comparing the lane change predicted trajectory with the actual trajectory. The results show that the absolute error of the lateral displacement of the predicted trajectory is concentrated between −0.3 m~0.3 m, the peak value of deflection angle velocity fluctuation is less than 2˚/s, and the acceleration is maintained in the range of 0~2 m/s2, which meets the rationality of the trajectory, lane change comfort and stability.

关 键 词：换道 运动学原理 序列二次规划 轨迹预测 

分 类 号：G63[文化科学—教育学]