基于二次规划的智能车辆动态换道轨迹规划研究

为实现智能车辆的自主换道操作并满足安全性、舒适性和实时性等约束条件，提出一种针对动态交通环境的换道轨迹规划模型。该模型由道路平面曲线表征模块、路径生成模块以及速度曲线生成模块组成。首先，在道路平面曲线表征模块中，模型基于实时获取的周边道路信息，利用切比雪夫多项式插值法回归拟合出连续可导的道路平面曲线函数，用以保证模型在各种道路平面线形上的普适性。然后，在路径生成模块中，根据换道车辆初始时刻的运动状态，建立一系列多项式方程，并利用牛顿迭代法求解方程未知参数，以此生成连接初始位置和目标位置的换道路径，用以保证换道轨迹的平滑性。最后，在速度曲线生成模块中，以满足防碰撞约束、跟驰加速度约束以及车辆运动状态约束为目标，构建二次规划模型，生成沿着换道路径的车辆速度曲线，用以保证换道轨迹的安全性和舒适性。此外，考虑到周边动态的交通环境，车辆系统在每个时间步内会循环调用提出的模型实时更新换道轨迹，直至车辆到达目标位置。仿真试验结果表明：应用提出的换道轨迹规划模型，车辆能够有效避免与周边动态车辆发生碰撞，成功完成换道；基于二次规划框架，模型优化求解时间明显缩短，满足轨迹规划的实时性和有效性要求。

Dynamic Lane-changing Trajectory Planning Model for Intelligent Vehicle Based on Quadratic Programming

To realize automated lane-changing maneuvers for intelligent vehicles and meet safety, comfort, and real-time requirements, a lane-changing trajectory planning model for dynamic environments was proposed. The model consisted of a road plane curve characterization module, a path generation module, and a velocity profile generation module. First, in the road plane curve characterization module, the Chebyshev polynomial interpolation method was used to fit the continuous and derivable surrounding road curve, which enabled the universality of the proposed model in various road plane alignments. Then, in the path generation module, according to the initial traffic states of the intelligent vehicle, a series of polynomial equations were established to generate a lane-changing path connecting the initial position and the candidate target positions, in which the equation parameters were solved using the Newton iteration method. Finally, in the velocity profile generation module, considering the collision avoidance, car-following acceleration, and vehicle motion state constraints, a quadratic programming problem was constructed to generate velocity profiles along the lane-changing path. Moreover, considering the surrounding dynamic traffic environments, the proposed model could be applied repeatedly in each time step to update the lane-changing trajectory until the intelligent vehicle reached the target position. The simulation results show that by employing the proposed lane-changing trajectory planning model, the intelligent vehicle can effectively avoid collisions with surrounding moving vehicles and successfully complete the lane-changing maneuver. Simultaneously, based on the quadratic programming framework, the model optimization time is significantly shortened, which meets the requirements for real-time planning and the effectiveness of trajectory planning.