城市道路汽车自动驾驶路径生成算法  

Trajectory Generation for Autonomous Driving Urban Vehicle

作　　者：赵倬祺 张引 张琨 王严 Zhao Zhuoqi;Zhang Yin;Zhang Kun;Wang Yan(Brilliance Auto R&D Center, Liaoning Shenyang 110141)

机构地区：[1]华晨汽车工程研究院,辽宁沈阳110141

出　　处：《汽车实用技术》2018年第4期38-40,共3页Automobile Applied Technology

基　　金：国家重点研发计划(2016YFB0101107)资助

摘　　要：文章通过离散化候选路径与参考路径之间的横向距离和车辆行驶过参考路径的弧长,形成多个目标状态。状态包括位置和方向。根据车辆当前状态和目标状态,求解五次多项式,使车辆与目标状态中间形成平滑连接。使用弗莱纳公式,将多项式上的点转换到参考路径平面坐标下,即形成候选路径。同时对平顺性、行驶偏差、路径长度计算路径代价,使用人工势场计算障碍物代价,选取最优路径。Target states,including positions and postures are generated by both discretization of distance between candidate trajectory and reference trajectory,and vehicle displacement represented by arc length of reference trajectory.Smooth candidate trajectories generated by quantic polynomial considering the start and end states,are transformed into scale-space by frenet frame.The optimal trajectory is selected by cost function involved jerk,driving deviation,trajectory length,besides the cost of obstacles calculated by artificial potential field.

关 键 词：人工势场 弗莱纳标架 代价函数 

分 类 号：U461.6[机械工程—车辆工程]