Abstract: At present, the autonomous driving system of vehicles has less involvement in curve road conditions. Based on Frenet coordinate system, aiming at the lane changing strategy of autonomous vehicles at curves, this paper proposes a trajectory-planning algorithm for handling dynamic obstacles with transverse and longitudinal decoupling. It solves the problem that the traditional coupling planning is difficult to take into account the curve fitting efficiency and safety. The transverse planning is based on the quasi-uniform cubic B-spline curve, while the longitudinal planning is based on dynamic planning and quadratic optimization. The model predictive control is used to track and control the lane changing trajectory. Finally, Matlab/Simulink simulation is used to verify that the proposed curve lane changing trajectory algorithm can change lanes smoothly and avoid dynamic obstacles safely.

Key words: Autonomous driving; Frenet coordinate; Curve lane changing; Trajectory planning; Trajectory tracking; Matlab/Simulink

摘要： 目前车辆的自动驾驶系统在弯道路况有较少的介入。基于 Frenet 坐标系，针对自动驾 驶汽车在弯道处的换道策略，提出了一种横纵向解耦的处理动态障碍物的轨迹规划算法，解 决了传统耦合规划难以兼顾曲线拟合效率与安全性。横向规划基于准均匀三次 B 样条曲线， 纵向规划则通过动态规划以及二次优化的思想。采用模型预测控制进行换道轨迹的跟踪与控 制。最终使用 Matlab/Simulink 仿真验证，所提出的弯道换道轨迹算法能平稳换道并安全避让 动态障碍物。

关键词: 自动驾驶；Frenet 坐标；弯道处换道；轨迹规划；轨迹跟踪；Matlab/Simulink