基于五次多项式的智能车辆轨迹规划

Trajectory planning of intelligent vehicle based on quintic polynomial

针对智能车辆换道轨迹中存在的侧向加速度过大或轨迹曲率不连续的问题,在对传统车辆换道模型进行比较分析的基础上,提出基于五次多项式换道模型的轨迹规划方法.基于换道轨迹安全性和效率性的要求,以车辆换道时的侧向加速度、换道时间和车辆横摆角速度为优化变量,设计目标函数.通过求解目标函数得到最优换道时间,进而得到最优换道轨迹.对等速偏移+正弦函数换道模型和五次多项式换道模型进行仿真对比.结果表明:利用五次多项式换道模型的轨迹规划方法,当路面附着系数为0.2时,侧向加速度最大值为0.45 m/s^(2),轨迹的曲率最大值为2.02×10^(-3)m^(-1);路面附着系数为0.6时,侧向加速度最大值为0.70 m/s^(2),轨迹的曲率最大值为1.12×10^(-3)m^(-1);路面附着系数为0.8时,侧向加速度最大值为0.81 m/s^(2),轨迹的曲率最大值为0.90×10^(-3)m^(-1),均小于等速偏移+正弦函数换道模型轨迹曲线的侧向加速度和曲率,验证了该模型的有效性.

To solve the problems of excessive lateral acceleration and discontinuous trajectory curvature in intelligent vehicle lane changing trajectories,the trajectory planning method based on the quintic polynomial lane changing model was proposed according to the comparative analysis of traditional vehicle lane changing models.Based on the requirements for safety and efficiency of lane changing trajectories,an objective function was designed with lateral acceleration,lane changing time and vehicle yaw rate as optimization variables.The optimal lane changing time was solved from the objective function to obtain the optimal lane changing trajectory.The simulation experiment was conducted to compare the constant velocity offset plus sine function lane changing model with the quintic polynomial lane changing model.The results show that according to the trajectory planning method of the quintic polynomial lane change model,when the road adhesion coefficient is 0.2,the maximum lateral acceleration is 0.45 m/s^(2)and the maximum curvature of the trajectory is 2.02×10^(-3)m^(-1),When the road adhesion coefficient is 0.6,the maximum lateral acceleration is 0.70 m/s^(2),and the maximum curvature value is 1.12×10^(-3)m^(-1) When the road adhesion coefficient is 0.8,the maximum lateral acceleration is 0.81 m/s^(2),and the maximum curvature value is 0.90×10^(-3)m^(-1),The acceleration and curvature of the trajectory curve of the lane change model are less than those of the constant velocity offset plus sine function,which can verify the effectiveness of the model.